{
  "version": 1,
  "asOf": "2026-09-19",
  "inputs": {
    "snapshots": [
      "2026-08-24",
      "2026-08-25",
      "2026-08-26",
      "2026-08-28",
      "2026-08-29",
      "2026-08-30",
      "2026-09-02",
      "2026-09-03",
      "2026-09-04",
      "2026-09-08",
      "2026-09-11",
      "2026-09-15",
      "2026-09-19"
    ],
    "changeset": null,
    "corrections": 18
  },
  "notes": [
    "No frontier movement is published for 2026-08-25 → 2026-08-26. The ranked board changed by 62% between those two snapshots — 56 slugs entered the ranked set and 44 left, of a 162-slug union — which is over the 20% ceiling this wire treats as a definition change rather than a market move. The frontier is a Pareto set over that board, so it moves when the board does. Price movement is unaffected: a per-model price delta does not depend on who else is ranked.",
    "13 corrections in docs/EMPIRICAL-GROUNDING.md carry no date stamp and are therefore not in the wire. An entry with no date cannot be placed on a timeline. The log itself is complete; this feed is the incomplete view of it."
  ],
  "events": [
    {
      "id": "frontier-join-2026-09-19-z-ai-glm-5-3",
      "date": "2026-09-19",
      "type": "frontier-join",
      "title": "Z.ai: GLM 5.3 joined the frontier",
      "summary": "1475.1 on LMArena at $1.40/M on 2026-09-19, against 1475.1 at $2.15/M on 2026-09-15. Model: z-ai/glm-5.3.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "z-ai/glm-5.3",
      "evidenceUrl": "/models/z-ai__glm-5.3/"
    },
    {
      "id": "frontier-leave-2026-09-19-google-gemma-3-4b-it",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "Google: Gemma 3 4B left the frontier",
      "summary": "1290.7 on LMArena at $0.0625/M on 2026-09-19, unchanged from 2026-09-15. Its own figures did not move — the board did. A frontier is a Pareto set, so a model leaves it when the rest of the board changes around it. The wire does not name what now beats it: the dominance rule refuses a replacement that costs capability, and the snapshot projection carries no context window or modalities to check that with. Model: google/gemma-3-4b-it.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "google/gemma-3-4b-it",
      "evidenceUrl": "/models/google__gemma-3-4b-it/"
    },
    {
      "id": "frontier-leave-2026-09-19-google-gemma-4-26b-a4b-it",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "Google: Gemma 4 26B A4B  left the frontier",
      "summary": "1434.5 on LMArena at $0.1425/M on 2026-09-19, unchanged from 2026-09-15. Its own figures did not move — the board did. A frontier is a Pareto set, so a model leaves it when the rest of the board changes around it. The wire does not name what now beats it: the dominance rule refuses a replacement that costs capability, and the snapshot projection carries no context window or modalities to check that with. Model: google/gemma-4-26b-a4b-it.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "google/gemma-4-26b-a4b-it",
      "evidenceUrl": "/models/google__gemma-4-26b-a4b-it/"
    },
    {
      "id": "frontier-leave-2026-09-19-google-gemma-4-31b-it",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "Google: Gemma 4 31B left the frontier",
      "summary": "1441.7 on LMArena at $0.1525/M on 2026-09-19, unchanged from 2026-09-15. Its own figures did not move — the board did. A frontier is a Pareto set, so a model leaves it when the rest of the board changes around it. The wire does not name what now beats it: the dominance rule refuses a replacement that costs capability, and the snapshot projection carries no context window or modalities to check that with. Model: google/gemma-4-31b-it.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "google/gemma-4-31b-it",
      "evidenceUrl": "/models/google__gemma-4-31b-it/"
    },
    {
      "id": "frontier-leave-2026-09-19-openai-gpt-oss-120b",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "OpenAI: gpt-oss-120b left the frontier",
      "summary": "1365.9 on LMArena at $0.2625/M on 2026-09-19, against 1365.9 at $0.0703/M on 2026-09-15. Model: openai/gpt-oss-120b.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "openai/gpt-oss-120b",
      "evidenceUrl": "/models/openai__gpt-oss-120b/"
    },
    {
      "id": "frontier-leave-2026-09-19-openai-gpt-oss-20b",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "OpenAI: gpt-oss-20b left the frontier",
      "summary": "1287.3 on LMArena at $0.055/M on 2026-09-19, unchanged from 2026-09-15. Its own figures did not move — the board did. A frontier is a Pareto set, so a model leaves it when the rest of the board changes around it. The wire does not name what now beats it: the dominance rule refuses a replacement that costs capability, and the snapshot projection carries no context window or modalities to check that with. Model: openai/gpt-oss-20b.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "openai/gpt-oss-20b",
      "evidenceUrl": "/models/openai__gpt-oss-20b/"
    },
    {
      "id": "frontier-leave-2026-09-19-qwen-qwen3-30b-a3b-instruct-2507",
      "date": "2026-09-19",
      "type": "frontier-leave",
      "title": "Qwen: Qwen3 30B A3B Instruct 2507 left the frontier",
      "summary": "1383.6 on LMArena at $0.0844/M on 2026-09-19, unchanged from 2026-09-15. Its own figures did not move — the board did. A frontier is a Pareto set, so a model leaves it when the rest of the board changes around it. The wire does not name what now beats it: the dominance rule refuses a replacement that costs capability, and the snapshot projection carries no context window or modalities to check that with. Model: qwen/qwen3-30b-a3b-instruct-2507.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-15 → 2026-09-19.",
      "modelSlug": "qwen/qwen3-30b-a3b-instruct-2507",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b-instruct-2507/"
    },
    {
      "id": "price-change-2026-09-19-deepseek-deepseek-v4-flash",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 is 55.6% cheaper on the balanced blend",
      "summary": "$0.1089/M on 2026-09-15, $0.0483/M on 2026-09-19 — a 55.6% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "price-change-2026-09-19-deepseek-deepseek-v4-flash-0731",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0731 is 33.3% cheaper on the balanced blend",
      "summary": "$0.075/M on 2026-09-15, $0.05/M on 2026-09-19 — a 33.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash-0731.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash-0731",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash-0731/"
    },
    {
      "id": "price-change-2026-09-19-deepseek-deepseek-v4-pro",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0423 is 73.6% cheaper on the balanced blend",
      "summary": "$2.00/M on 2026-09-15, $0.5279/M on 2026-09-19 — a 73.6% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro/"
    },
    {
      "id": "price-change-2026-09-19-deepseek-deepseek-v4-pro-0813",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 12.2% cheaper on the balanced blend",
      "summary": "$0.988/M on 2026-09-15, $0.8672/M on 2026-09-19 — a 12.2% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-09-19-gryphe-mythomax-l2-13b",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "MythoMax 13B is 45.8% dearer on the balanced blend",
      "summary": "$0.06/M on 2026-09-15, $0.0875/M on 2026-09-19 — a 45.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: gryphe/mythomax-l2-13b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "gryphe/mythomax-l2-13b",
      "evidenceUrl": "/models/gryphe__mythomax-l2-13b/"
    },
    {
      "id": "price-change-2026-09-19-minimax-minimax-m1",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "MiniMax: MiniMax M1 is 11.7% cheaper on the balanced blend",
      "summary": "$0.9625/M on 2026-09-15, $0.85/M on 2026-09-19 — a 11.7% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: minimax/minimax-m1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "minimax/minimax-m1",
      "evidenceUrl": "/models/minimax__minimax-m1/"
    },
    {
      "id": "price-change-2026-09-19-mistralai-mistral-small-3-2-24b-instruct",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Mistral: Mistral Small 3.2 24B is 25% dearer on the balanced blend",
      "summary": "$0.1062/M on 2026-09-15, $0.1328/M on 2026-09-19 — a 25% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: mistralai/mistral-small-3.2-24b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "mistralai/mistral-small-3.2-24b-instruct",
      "evidenceUrl": "/models/mistralai__mistral-small-3.2-24b-instruct/"
    },
    {
      "id": "price-change-2026-09-19-moonshotai-kimi-k2-7-code",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "MoonshotAI: Kimi K2.7 Code is 5.4% cheaper on the balanced blend",
      "summary": "$1.41/M on 2026-09-15, $1.33/M on 2026-09-19 — a 5.4% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: moonshotai/kimi-k2.7-code.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "moonshotai/kimi-k2.7-code",
      "evidenceUrl": "/models/moonshotai__kimi-k2.7-code/"
    },
    {
      "id": "price-change-2026-09-19-moonshotai-kimi-k3",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "MoonshotAI: Kimi K3 is 35.9% cheaper on the balanced blend",
      "summary": "$5.31/M on 2026-09-15, $3.40/M on 2026-09-19 — a 35.9% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: moonshotai/kimi-k3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "moonshotai/kimi-k3",
      "evidenceUrl": "/models/moonshotai__kimi-k3/"
    },
    {
      "id": "price-change-2026-09-19-nvidia-nemotron-3-5-lightning",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3.5 Lightning is 6.8% cheaper on the balanced blend",
      "summary": "$0.11/M on 2026-09-15, $0.1025/M on 2026-09-19 — a 6.8% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3.5-lightning.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3.5-lightning",
      "evidenceUrl": "/models/nvidia__nemotron-3.5-lightning/"
    },
    {
      "id": "price-change-2026-09-19-nvidia-nemotron-3-nano-30b-a3b",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Nano 30B A3B is 20% dearer on the balanced blend",
      "summary": "$0.0875/M on 2026-09-15, $0.105/M on 2026-09-19 — a 20% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-nano-30b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-nano-30b-a3b",
      "evidenceUrl": "/models/nvidia__nemotron-3-nano-30b-a3b/"
    },
    {
      "id": "price-change-2026-09-19-openai-gpt-oss-120b",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "OpenAI: gpt-oss-120b is 273.4% dearer on the balanced blend",
      "summary": "$0.0703/M on 2026-09-15, $0.2625/M on 2026-09-19 — a 273.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: openai/gpt-oss-120b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "openai/gpt-oss-120b",
      "evidenceUrl": "/models/openai__gpt-oss-120b/"
    },
    {
      "id": "price-change-2026-09-19-qwen-qwen3-vl-30b-a3b-instruct",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Qwen: Qwen3 VL 30B A3B Instruct is 13.3% cheaper on the balanced blend",
      "summary": "$0.2625/M on 2026-09-15, $0.2275/M on 2026-09-19 — a 13.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-vl-30b-a3b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-vl-30b-a3b-instruct",
      "evidenceUrl": "/models/qwen__qwen3-vl-30b-a3b-instruct/"
    },
    {
      "id": "price-change-2026-09-19-tencent-hy3",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Tencent: Hy3 is 37.5% cheaper on the balanced blend",
      "summary": "$0.231/M on 2026-09-15, $0.1444/M on 2026-09-19 — a 37.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: tencent/hy3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-19-z-ai-glm-5-2",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Z.ai: GLM 5.2 is 60.4% cheaper on the balanced blend",
      "summary": "$2.15/M on 2026-09-15, $0.8514/M on 2026-09-19 — a 60.4% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.2.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "price-change-2026-09-19-z-ai-glm-5-3",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Z.ai: GLM 5.3 is 35% cheaper on the balanced blend",
      "summary": "$2.15/M on 2026-09-15, $1.40/M on 2026-09-19 — a 35% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.3",
      "evidenceUrl": "/models/z-ai__glm-5.3/"
    },
    {
      "id": "price-change-2026-09-19-z-ai-glm-5-3-flash",
      "date": "2026-09-19",
      "type": "price-change",
      "title": "Z.ai: GLM 5.3 Flash is 40% cheaper on the balanced blend",
      "summary": "$0.2375/M on 2026-09-15, $0.1425/M on 2026-09-19 — a 40% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.3-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-15 → 2026-09-19. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.3-flash",
      "evidenceUrl": "/models/z-ai__glm-5.3-flash/"
    },
    {
      "id": "correction-2026-09-19-48-the-model-page-still-flattened-the-ladder-that-17-un-flattened",
      "date": "2026-09-19",
      "type": "correction",
      "title": "Correction: the model page still flattened the ladder that §17 un-flattened",
      "summary": "Logged 2026-09-19 as §48 of docs/EMPIRICAL-GROUNDING.md. The wire carries the fact that a correction was published; the log carries what was claimed, what is true, and how it was found. Corrections are not summarised here — shortening one is how a load-bearing qualification gets dropped. A vendor changing a price is not us being wrong; this is.",
      "conditions": "Correction log, docs/EMPIRICAL-GROUNDING.md §48, logged 2026-09-19.",
      "evidenceUrl": "/corrections/"
    },
    {
      "id": "frontier-join-2026-09-15-deepseek-deepseek-v4-flash",
      "date": "2026-09-15",
      "type": "frontier-join",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 joined the frontier",
      "summary": "1431.8 on LMArena at $0.1089/M on 2026-09-15, against 1431.9 at $0.1071/M on 2026-09-11. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-11 → 2026-09-15.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "frontier-join-2026-09-15-qwen-qwen3-30b-a3b-instruct-2507",
      "date": "2026-09-15",
      "type": "frontier-join",
      "title": "Qwen: Qwen3 30B A3B Instruct 2507 joined the frontier",
      "summary": "1383.6 on LMArena at $0.0844/M on 2026-09-15, against 1383.5 at $0.1425/M on 2026-09-11. Model: qwen/qwen3-30b-a3b-instruct-2507.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-11 → 2026-09-15.",
      "modelSlug": "qwen/qwen3-30b-a3b-instruct-2507",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b-instruct-2507/"
    },
    {
      "id": "frontier-leave-2026-09-15-tencent-hy3",
      "date": "2026-09-15",
      "type": "frontier-leave",
      "title": "Tencent: Hy3 left the frontier",
      "summary": "1440.6 on LMArena at $0.231/M on 2026-09-15, against 1439.9 at $0.1444/M on 2026-09-11. Model: tencent/hy3.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-11 → 2026-09-15.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-15-deepseek-deepseek-v4-flash-0731",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0731 is 20% cheaper on the balanced blend",
      "summary": "$0.0938/M on 2026-09-11, $0.075/M on 2026-09-15 — a 20% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash-0731.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash-0731",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash-0731/"
    },
    {
      "id": "price-change-2026-09-15-deepseek-deepseek-v4-pro",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0423 is 68.8% dearer on the balanced blend",
      "summary": "$1.18/M on 2026-09-11, $2.00/M on 2026-09-15 — a 68.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro/"
    },
    {
      "id": "price-change-2026-09-15-deepseek-deepseek-v4-pro-0813",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 13.7% dearer on the balanced blend",
      "summary": "$0.8692/M on 2026-09-11, $0.988/M on 2026-09-15 — a 13.7% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-09-15-google-gemma-4-26b-a4b-it",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Google: Gemma 4 26B A4B  is 64.7% dearer on the balanced blend",
      "summary": "$0.0865/M on 2026-09-11, $0.1425/M on 2026-09-15 — a 64.7% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: google/gemma-4-26b-a4b-it.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "google/gemma-4-26b-a4b-it",
      "evidenceUrl": "/models/google__gemma-4-26b-a4b-it/"
    },
    {
      "id": "price-change-2026-09-15-meta-llama-llama-4-maverick",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Meta: Llama 4 Maverick is 6.3% cheaper on the balanced blend",
      "summary": "$0.324/M on 2026-09-11, $0.3037/M on 2026-09-15 — a 6.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-4-maverick.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-4-maverick",
      "evidenceUrl": "/models/meta-llama__llama-4-maverick/"
    },
    {
      "id": "price-change-2026-09-15-meta-muse-glimmer-30b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Meta: Muse Glimmer 30B is 27.5% dearer on the balanced blend",
      "summary": "$0.5/M on 2026-09-11, $0.6375/M on 2026-09-15 — a 27.5% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta/muse-glimmer-30b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta/muse-glimmer-30b",
      "evidenceUrl": "/models/meta__muse-glimmer-30b/"
    },
    {
      "id": "price-change-2026-09-15-moonshotai-kimi-k3",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "MoonshotAI: Kimi K3 is 13.4% dearer on the balanced blend",
      "summary": "$4.68/M on 2026-09-11, $5.31/M on 2026-09-15 — a 13.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: moonshotai/kimi-k3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "moonshotai/kimi-k3",
      "evidenceUrl": "/models/moonshotai__kimi-k3/"
    },
    {
      "id": "price-change-2026-09-15-nvidia-nemotron-3-super-120b-a12b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Super is 5.3% dearer on the balanced blend",
      "summary": "$0.1638/M on 2026-09-11, $0.1725/M on 2026-09-15 — a 5.3% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-super-120b-a12b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-super-120b-a12b",
      "evidenceUrl": "/models/nvidia__nemotron-3-super-120b-a12b/"
    },
    {
      "id": "price-change-2026-09-15-nvidia-nemotron-3-ultra-550b-a55b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Ultra is 16% cheaper on the balanced blend",
      "summary": "$1.25/M on 2026-09-11, $1.05/M on 2026-09-15 — a 16% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-ultra-550b-a55b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-ultra-550b-a55b",
      "evidenceUrl": "/models/nvidia__nemotron-3-ultra-550b-a55b/"
    },
    {
      "id": "price-change-2026-09-15-qwen-qwen3-14b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Qwen: Qwen3 14B is 62.3% cheaper on the balanced blend",
      "summary": "$0.3981/M on 2026-09-11, $0.15/M on 2026-09-15 — a 62.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-14b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-14b",
      "evidenceUrl": "/models/qwen__qwen3-14b/"
    },
    {
      "id": "price-change-2026-09-15-qwen-qwen3-235b-a22b-2507",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Qwen: Qwen3 235B A22B Instruct 2507 is 60.2% cheaper on the balanced blend",
      "summary": "$0.385/M on 2026-09-11, $0.1531/M on 2026-09-15 — a 60.2% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-235b-a22b-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-235b-a22b-2507",
      "evidenceUrl": "/models/qwen__qwen3-235b-a22b-2507/"
    },
    {
      "id": "price-change-2026-09-15-qwen-qwen3-30b-a3b-instruct-2507",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Qwen: Qwen3 30B A3B Instruct 2507 is 40.8% cheaper on the balanced blend",
      "summary": "$0.1425/M on 2026-09-11, $0.0844/M on 2026-09-15 — a 40.8% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-30b-a3b-instruct-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-30b-a3b-instruct-2507",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b-instruct-2507/"
    },
    {
      "id": "price-change-2026-09-15-qwen-qwen3-5-35b-a3b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Qwen: Qwen3.5-35B-A3B is 18.3% cheaper on the balanced blend",
      "summary": "$0.5469/M on 2026-09-11, $0.4469/M on 2026-09-15 — a 18.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-35b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-35b-a3b",
      "evidenceUrl": "/models/qwen__qwen3.5-35b-a3b/"
    },
    {
      "id": "price-change-2026-09-15-qwen-qwen3-8-27b",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Qwen: Qwen3.8 27B is 25.1% cheaper on the balanced blend",
      "summary": "$1.06/M on 2026-09-11, $0.798/M on 2026-09-15 — a 25.1% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.8-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.8-27b",
      "evidenceUrl": "/models/qwen__qwen3.8-27b/"
    },
    {
      "id": "price-change-2026-09-15-tencent-hy3",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Tencent: Hy3 is 60% dearer on the balanced blend",
      "summary": "$0.1444/M on 2026-09-11, $0.231/M on 2026-09-15 — a 60% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: tencent/hy3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-15-z-ai-glm-5-2",
      "date": "2026-09-15",
      "type": "price-change",
      "title": "Z.ai: GLM 5.2 is 126.3% dearer on the balanced blend",
      "summary": "$0.95/M on 2026-09-11, $2.15/M on 2026-09-15 — a 126.3% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.2.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-11 → 2026-09-15. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "frontier-join-2026-09-11-anthropic-claude-fable-5-1",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "Anthropic: Claude Fable 5.1 joined the frontier",
      "summary": "1513.7 on LMArena at $20.00/M on 2026-09-11, against unrated at $20.00/M on 2026-09-08. Model: anthropic/claude-fable-5.1.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "anthropic/claude-fable-5.1",
      "evidenceUrl": "/models/anthropic__claude-fable-5.1/"
    },
    {
      "id": "frontier-join-2026-09-11-google-gemini-3-8-flash",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "Google: Gemini 3.8 Flash joined the frontier",
      "summary": "1495.2 on LMArena at $1.50/M on 2026-09-11, against unrated at $1.50/M on 2026-09-08. Model: google/gemini-3.8-flash.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "google/gemini-3.8-flash",
      "evidenceUrl": "/models/google__gemini-3.8-flash/"
    },
    {
      "id": "frontier-join-2026-09-11-google-gemma-3-4b-it",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "Google: Gemma 3 4B joined the frontier",
      "summary": "1290.6 on LMArena at $0.0625/M on 2026-09-11, against 1290.8 at $0.0625/M on 2026-09-08. Model: google/gemma-3-4b-it.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "google/gemma-3-4b-it",
      "evidenceUrl": "/models/google__gemma-3-4b-it/"
    },
    {
      "id": "frontier-join-2026-09-11-openai-gpt-oss-120b",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "OpenAI: gpt-oss-120b joined the frontier",
      "summary": "1365.9 on LMArena at $0.0703/M on 2026-09-11, against 1365.6 at $0.0703/M on 2026-09-08. Model: openai/gpt-oss-120b.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "openai/gpt-oss-120b",
      "evidenceUrl": "/models/openai__gpt-oss-120b/"
    },
    {
      "id": "frontier-join-2026-09-11-openai-gpt-oss-20b",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "OpenAI: gpt-oss-20b joined the frontier",
      "summary": "1287.2 on LMArena at $0.055/M on 2026-09-11, against 1287.8 at $0.055/M on 2026-09-08. Model: openai/gpt-oss-20b.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "openai/gpt-oss-20b",
      "evidenceUrl": "/models/openai__gpt-oss-20b/"
    },
    {
      "id": "frontier-join-2026-09-11-z-ai-glm-5-3-flash",
      "date": "2026-09-11",
      "type": "frontier-join",
      "title": "Z.ai: GLM 5.3 Flash joined the frontier",
      "summary": "1470.8 on LMArena at $0.2375/M on 2026-09-11, against unrated at $0.1128/M on 2026-09-08. Model: z-ai/glm-5.3-flash.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "z-ai/glm-5.3-flash",
      "evidenceUrl": "/models/z-ai__glm-5.3-flash/"
    },
    {
      "id": "frontier-leave-2026-09-11-deepseek-deepseek-v4-flash",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 left the frontier",
      "summary": "1431.9 on LMArena at $0.1071/M on 2026-09-11, against 1431.6 at $0.1069/M on 2026-09-08. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "frontier-leave-2026-09-11-google-gemini-3-7-flash",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "Google: Gemini 3.7 Flash left the frontier",
      "summary": "1491.4 on LMArena at $1.50/M on 2026-09-11, against 1490.2 at $1.50/M on 2026-09-08. Model: google/gemini-3.7-flash.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "google/gemini-3.7-flash",
      "evidenceUrl": "/models/google__gemini-3.7-flash/"
    },
    {
      "id": "frontier-leave-2026-09-11-qwen-qwen3-30b-a3b-instruct-2507",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "Qwen: Qwen3 30B A3B Instruct 2507 left the frontier",
      "summary": "1383.5 on LMArena at $0.1425/M on 2026-09-11, against 1384.3 at $0.0844/M on 2026-09-08. Model: qwen/qwen3-30b-a3b-instruct-2507.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "qwen/qwen3-30b-a3b-instruct-2507",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b-instruct-2507/"
    },
    {
      "id": "frontier-leave-2026-09-11-upstage-solar-pro4",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "Upstage: Solar Pro 4 left the frontier",
      "summary": "1377.5 on LMArena at $0.1575/M on 2026-09-11, against 1376.2 at $0.0525/M on 2026-09-08. Model: upstage/solar-pro4.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "upstage/solar-pro4",
      "evidenceUrl": "/models/upstage__solar-pro4/"
    },
    {
      "id": "frontier-leave-2026-09-11-xiaomi-mimo-v2-5-pro",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "Xiaomi: MiMo-V2.5-Pro left the frontier",
      "summary": "1465.3 on LMArena at $0.5437/M on 2026-09-11, against 1465 at $0.5437/M on 2026-09-08. Model: xiaomi/mimo-v2.5-pro.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "xiaomi/mimo-v2.5-pro",
      "evidenceUrl": "/models/xiaomi__mimo-v2.5-pro/"
    },
    {
      "id": "frontier-leave-2026-09-11-z-ai-glm-5-2",
      "date": "2026-09-11",
      "type": "frontier-leave",
      "title": "Z.ai: GLM 5.2 left the frontier",
      "summary": "1466.7 on LMArena at $0.95/M on 2026-09-11, against 1465.4 at $1.48/M on 2026-09-08. Model: z-ai/glm-5.2.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-08 → 2026-09-11.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "price-change-2026-09-11-deepseek-deepseek-chat-v3-0324",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V3 0324 is 12.9% cheaper on the balanced blend",
      "summary": "$0.5025/M on 2026-09-08, $0.4375/M on 2026-09-11 — a 12.9% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-chat-v3-0324.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-chat-v3-0324",
      "evidenceUrl": "/models/deepseek__deepseek-chat-v3-0324/"
    },
    {
      "id": "price-change-2026-09-11-google-gemma-4-26b-a4b-it",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Google: Gemma 4 26B A4B  is 37.1% cheaper on the balanced blend",
      "summary": "$0.1375/M on 2026-09-08, $0.0865/M on 2026-09-11 — a 37.1% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: google/gemma-4-26b-a4b-it.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "google/gemma-4-26b-a4b-it",
      "evidenceUrl": "/models/google__gemma-4-26b-a4b-it/"
    },
    {
      "id": "price-change-2026-09-11-moonshotai-kimi-k3",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "MoonshotAI: Kimi K3 is 22% cheaper on the balanced blend",
      "summary": "$6.00/M on 2026-09-08, $4.68/M on 2026-09-11 — a 22% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: moonshotai/kimi-k3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "moonshotai/kimi-k3",
      "evidenceUrl": "/models/moonshotai__kimi-k3/"
    },
    {
      "id": "price-change-2026-09-11-qwen-qwen3-235b-a22b-2507",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Qwen: Qwen3 235B A22B Instruct 2507 is 87.8% dearer on the balanced blend",
      "summary": "$0.205/M on 2026-09-08, $0.385/M on 2026-09-11 — a 87.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-235b-a22b-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-235b-a22b-2507",
      "evidenceUrl": "/models/qwen__qwen3-235b-a22b-2507/"
    },
    {
      "id": "price-change-2026-09-11-qwen-qwen3-30b-a3b-instruct-2507",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Qwen: Qwen3 30B A3B Instruct 2507 is 68.8% dearer on the balanced blend",
      "summary": "$0.0844/M on 2026-09-08, $0.1425/M on 2026-09-11 — a 68.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-30b-a3b-instruct-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-30b-a3b-instruct-2507",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b-instruct-2507/"
    },
    {
      "id": "price-change-2026-09-11-qwen-qwen3-5-122b-a10b",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Qwen: Qwen3.5-122B-A10B is 12.5% cheaper on the balanced blend",
      "summary": "$0.8175/M on 2026-09-08, $0.715/M on 2026-09-11 — a 12.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-122b-a10b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-122b-a10b",
      "evidenceUrl": "/models/qwen__qwen3.5-122b-a10b/"
    },
    {
      "id": "price-change-2026-09-11-upstage-solar-pro4",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Upstage: Solar Pro 4 is 200% dearer on the balanced blend",
      "summary": "$0.0525/M on 2026-09-08, $0.1575/M on 2026-09-11 — a 200% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: upstage/solar-pro4.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "upstage/solar-pro4",
      "evidenceUrl": "/models/upstage__solar-pro4/"
    },
    {
      "id": "price-change-2026-09-11-z-ai-glm-4-6",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Z.ai: GLM 4.6 is 21% cheaper on the balanced blend",
      "summary": "$0.9625/M on 2026-09-08, $0.76/M on 2026-09-11 — a 21% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-4.6.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-4.6",
      "evidenceUrl": "/models/z-ai__glm-4.6/"
    },
    {
      "id": "price-change-2026-09-11-z-ai-glm-5-2",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Z.ai: GLM 5.2 is 36% cheaper on the balanced blend",
      "summary": "$1.48/M on 2026-09-08, $0.95/M on 2026-09-11 — a 36% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.2.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "price-change-2026-09-11-z-ai-glm-5-3-flash",
      "date": "2026-09-11",
      "type": "price-change",
      "title": "Z.ai: GLM 5.3 Flash is 110.5% dearer on the balanced blend",
      "summary": "$0.1128/M on 2026-09-08, $0.2375/M on 2026-09-11 — a 110.5% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.3-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-08 → 2026-09-11. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.3-flash",
      "evidenceUrl": "/models/z-ai__glm-5.3-flash/"
    },
    {
      "id": "price-change-2026-09-08-deepseek-deepseek-chat-v3-0324",
      "date": "2026-09-08",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V3 0324 is 14.9% dearer on the balanced blend",
      "summary": "$0.4375/M on 2026-09-04, $0.5025/M on 2026-09-08 — a 14.9% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-chat-v3-0324.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-04 → 2026-09-08. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-chat-v3-0324",
      "evidenceUrl": "/models/deepseek__deepseek-chat-v3-0324/"
    },
    {
      "id": "price-change-2026-09-08-deepseek-deepseek-chat-v3-1",
      "date": "2026-09-08",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V3.1 is 48.5% cheaper on the balanced blend",
      "summary": "$0.825/M on 2026-09-04, $0.425/M on 2026-09-08 — a 48.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-chat-v3.1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-04 → 2026-09-08. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-chat-v3.1",
      "evidenceUrl": "/models/deepseek__deepseek-chat-v3.1/"
    },
    {
      "id": "price-change-2026-09-08-qwen-qwen3-14b",
      "date": "2026-09-08",
      "type": "price-change",
      "title": "Qwen: Qwen3 14B is 165.4% dearer on the balanced blend",
      "summary": "$0.15/M on 2026-09-04, $0.3981/M on 2026-09-08 — a 165.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-14b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-04 → 2026-09-08. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-14b",
      "evidenceUrl": "/models/qwen__qwen3-14b/"
    },
    {
      "id": "price-change-2026-09-08-qwen-qwen3-235b-a22b-2507",
      "date": "2026-09-08",
      "type": "price-change",
      "title": "Qwen: Qwen3 235B A22B Instruct 2507 is 33.9% dearer on the balanced blend",
      "summary": "$0.1531/M on 2026-09-04, $0.205/M on 2026-09-08 — a 33.9% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-235b-a22b-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-04 → 2026-09-08. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-235b-a22b-2507",
      "evidenceUrl": "/models/qwen__qwen3-235b-a22b-2507/"
    },
    {
      "id": "price-change-2026-09-08-qwen-qwen3-5-35b-a3b",
      "date": "2026-09-08",
      "type": "price-change",
      "title": "Qwen: Qwen3.5-35B-A3B is 121% dearer on the balanced blend",
      "summary": "$0.2475/M on 2026-09-04, $0.5469/M on 2026-09-08 — a 121% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-35b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-04 → 2026-09-08. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-35b-a3b",
      "evidenceUrl": "/models/qwen__qwen3.5-35b-a3b/"
    },
    {
      "id": "price-change-2026-09-04-deepseek-deepseek-v4-flash",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 is 10.6% dearer on the balanced blend",
      "summary": "$0.0992/M on 2026-09-03, $0.1097/M on 2026-09-04 — a 10.6% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "price-change-2026-09-04-deepseek-deepseek-v4-pro-0813",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 12.2% cheaper on the balanced blend",
      "summary": "$0.99/M on 2026-09-03, $0.8692/M on 2026-09-04 — a 12.2% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-09-04-nvidia-nemotron-3-ultra-550b-a55b",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Ultra is 19% dearer on the balanced blend",
      "summary": "$1.05/M on 2026-09-03, $1.25/M on 2026-09-04 — a 19% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-ultra-550b-a55b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-ultra-550b-a55b",
      "evidenceUrl": "/models/nvidia__nemotron-3-ultra-550b-a55b/"
    },
    {
      "id": "price-change-2026-09-04-qwen-qwen3-5-35b-a3b",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "Qwen: Qwen3.5-35B-A3B is 50.5% cheaper on the balanced blend",
      "summary": "$0.5/M on 2026-09-03, $0.2475/M on 2026-09-04 — a 50.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-35b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-35b-a3b",
      "evidenceUrl": "/models/qwen__qwen3.5-35b-a3b/"
    },
    {
      "id": "price-change-2026-09-04-qwen-qwen3-6-27b",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "Qwen: Qwen3.6 27B is 46.3% cheaper on the balanced blend",
      "summary": "$1.35/M on 2026-09-03, $0.725/M on 2026-09-04 — a 46.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.6-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.6-27b",
      "evidenceUrl": "/models/qwen__qwen3.6-27b/"
    },
    {
      "id": "price-change-2026-09-04-undi95-remm-slerp-l2-13b",
      "date": "2026-09-04",
      "type": "price-change",
      "title": "ReMM SLERP 13B is 15% cheaper on the balanced blend",
      "summary": "$0.5/M on 2026-09-03, $0.425/M on 2026-09-04 — a 15% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: undi95/remm-slerp-l2-13b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-03 → 2026-09-04. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "undi95/remm-slerp-l2-13b",
      "evidenceUrl": "/models/undi95__remm-slerp-l2-13b/"
    },
    {
      "id": "frontier-join-2026-09-03-tencent-hy3",
      "date": "2026-09-03",
      "type": "frontier-join",
      "title": "Tencent: Hy3 joined the frontier",
      "summary": "1441.2 on LMArena at $0.1444/M on 2026-09-03, against 1441.2 at $0.231/M on 2026-09-02. Model: tencent/hy3.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-09-02 → 2026-09-03.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-03-deepseek-deepseek-chat-v3-1",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V3.1 is 94.1% dearer on the balanced blend",
      "summary": "$0.425/M on 2026-09-02, $0.825/M on 2026-09-03 — a 94.1% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-chat-v3.1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-chat-v3.1",
      "evidenceUrl": "/models/deepseek__deepseek-chat-v3.1/"
    },
    {
      "id": "price-change-2026-09-03-deepseek-deepseek-v4-flash",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 is 10.5% cheaper on the balanced blend",
      "summary": "$0.1108/M on 2026-09-02, $0.0992/M on 2026-09-03 — a 10.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "price-change-2026-09-03-deepseek-deepseek-v4-pro-0813",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 40.8% cheaper on the balanced blend",
      "summary": "$1.67/M on 2026-09-02, $0.99/M on 2026-09-03 — a 40.8% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-09-03-meta-llama-llama-3-3-70b-instruct",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Meta: Llama 3.3 70B Instruct is 78.2% cheaper on the balanced blend",
      "summary": "$0.71/M on 2026-09-02, $0.155/M on 2026-09-03 — a 78.2% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-3.3-70b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-3.3-70b-instruct",
      "evidenceUrl": "/models/meta-llama__llama-3.3-70b-instruct/"
    },
    {
      "id": "price-change-2026-09-03-nvidia-nemotron-3-ultra-550b-a55b",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Ultra is 16% cheaper on the balanced blend",
      "summary": "$1.25/M on 2026-09-02, $1.05/M on 2026-09-03 — a 16% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-ultra-550b-a55b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-ultra-550b-a55b",
      "evidenceUrl": "/models/nvidia__nemotron-3-ultra-550b-a55b/"
    },
    {
      "id": "price-change-2026-09-03-qwen-qwen2-5-vl-72b-instruct",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Qwen: Qwen2.5 VL 72B Instruct is 126.7% dearer on the balanced blend",
      "summary": "$0.375/M on 2026-09-02, $0.85/M on 2026-09-03 — a 126.7% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen2.5-vl-72b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen2.5-vl-72b-instruct",
      "evidenceUrl": "/models/qwen__qwen2.5-vl-72b-instruct/"
    },
    {
      "id": "price-change-2026-09-03-qwen-qwen3-5-397b-a17b",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Qwen: Qwen3.5 397B A17B is 46.7% dearer on the balanced blend",
      "summary": "$0.8775/M on 2026-09-02, $1.29/M on 2026-09-03 — a 46.7% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-397b-a17b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-397b-a17b",
      "evidenceUrl": "/models/qwen__qwen3.5-397b-a17b/"
    },
    {
      "id": "price-change-2026-09-03-qwen-qwen3-8-27b",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Qwen: Qwen3.8 27B is 11.4% dearer on the balanced blend",
      "summary": "$0.9562/M on 2026-09-02, $1.06/M on 2026-09-03 — a 11.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.8-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.8-27b",
      "evidenceUrl": "/models/qwen__qwen3.8-27b/"
    },
    {
      "id": "price-change-2026-09-03-tencent-hy3",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Tencent: Hy3 is 37.5% cheaper on the balanced blend",
      "summary": "$0.231/M on 2026-09-02, $0.1444/M on 2026-09-03 — a 37.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: tencent/hy3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-03-z-ai-glm-4-6",
      "date": "2026-09-03",
      "type": "price-change",
      "title": "Z.ai: GLM 4.6 is 26.6% dearer on the balanced blend",
      "summary": "$0.76/M on 2026-09-02, $0.9625/M on 2026-09-03 — a 26.6% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-4.6.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-09-02 → 2026-09-03. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-4.6",
      "evidenceUrl": "/models/z-ai__glm-4.6/"
    },
    {
      "id": "frontier-join-2026-09-02-z-ai-glm-5-2",
      "date": "2026-09-02",
      "type": "frontier-join",
      "title": "Z.ai: GLM 5.2 joined the frontier",
      "summary": "1465.4 on LMArena at $1.48/M on 2026-09-02, against 1465.4 at $1.83/M on 2026-08-30. Model: z-ai/glm-5.2.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-08-30 → 2026-09-02.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "frontier-leave-2026-09-02-tencent-hy3",
      "date": "2026-09-02",
      "type": "frontier-leave",
      "title": "Tencent: Hy3 left the frontier",
      "summary": "1441.2 on LMArena at $0.231/M on 2026-09-02, against 1441.2 at $0.1444/M on 2026-08-30. Model: tencent/hy3.",
      "conditions": "Default board: standard delivery, not retired, LMArena general axis, balanced-workload blend. Rank snapshots 2026-08-30 → 2026-09-02.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-02-anthracite-org-magnum-v4-72b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Magnum v4 72B is 10.7% cheaper on the balanced blend",
      "summary": "$3.50/M on 2026-08-30, $3.13/M on 2026-09-02 — a 10.7% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: anthracite-org/magnum-v4-72b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "anthracite-org/magnum-v4-72b",
      "evidenceUrl": "/models/anthracite-org__magnum-v4-72b/"
    },
    {
      "id": "price-change-2026-09-02-deepseek-deepseek-chat-v3-1",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V3.1 is 48.5% cheaper on the balanced blend",
      "summary": "$0.825/M on 2026-08-30, $0.425/M on 2026-09-02 — a 48.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-chat-v3.1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-chat-v3.1",
      "evidenceUrl": "/models/deepseek__deepseek-chat-v3.1/"
    },
    {
      "id": "price-change-2026-09-02-deepseek-deepseek-v4-flash",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 is 9.4% dearer on the balanced blend",
      "summary": "$0.1013/M on 2026-08-30, $0.1108/M on 2026-09-02 — a 9.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "price-change-2026-09-02-deepseek-deepseek-v4-flash-0731",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0731 is 25.1% dearer on the balanced blend",
      "summary": "$0.075/M on 2026-08-30, $0.0938/M on 2026-09-02 — a 25.1% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash-0731.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash-0731",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash-0731/"
    },
    {
      "id": "price-change-2026-09-02-deepseek-deepseek-v4-pro",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0423 is 19.8% dearer on the balanced blend",
      "summary": "$1.09/M on 2026-08-30, $1.30/M on 2026-09-02 — a 19.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro/"
    },
    {
      "id": "price-change-2026-09-02-deepseek-deepseek-v4-pro-0813",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 15.5% cheaper on the balanced blend",
      "summary": "$1.98/M on 2026-08-30, $1.67/M on 2026-09-02 — a 15.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-09-02-google-gemini-3-7-flash",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Google: Gemini 3.7 Flash is 100% dearer on the balanced blend",
      "summary": "$0.75/M on 2026-08-30, $1.50/M on 2026-09-02 — a 100% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: google/gemini-3.7-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "google/gemini-3.7-flash",
      "evidenceUrl": "/models/google__gemini-3.7-flash/"
    },
    {
      "id": "price-change-2026-09-02-mancer-weaver",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Mancer: Weaver (alpha) is 13.3% cheaper on the balanced blend",
      "summary": "$0.5625/M on 2026-08-30, $0.4875/M on 2026-09-02 — a 13.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: mancer/weaver.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "mancer/weaver",
      "evidenceUrl": "/models/mancer__weaver/"
    },
    {
      "id": "price-change-2026-09-02-meta-llama-llama-4-maverick",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Meta: Llama 4 Maverick is 7.4% cheaper on the balanced blend",
      "summary": "$0.35/M on 2026-08-30, $0.324/M on 2026-09-02 — a 7.4% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-4-maverick.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-4-maverick",
      "evidenceUrl": "/models/meta-llama__llama-4-maverick/"
    },
    {
      "id": "price-change-2026-09-02-meta-llama-llama-4-scout",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Meta: Llama 4 Scout is 10.4% cheaper on the balanced blend",
      "summary": "$0.1675/M on 2026-08-30, $0.15/M on 2026-09-02 — a 10.4% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-4-scout.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-4-scout",
      "evidenceUrl": "/models/meta-llama__llama-4-scout/"
    },
    {
      "id": "price-change-2026-09-02-meta-muse-glimmer-30b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Meta: Muse Glimmer 30B is 17.6% cheaper on the balanced blend",
      "summary": "$0.6375/M on 2026-08-30, $0.525/M on 2026-09-02 — a 17.6% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta/muse-glimmer-30b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta/muse-glimmer-30b",
      "evidenceUrl": "/models/meta__muse-glimmer-30b/"
    },
    {
      "id": "price-change-2026-09-02-mistralai-devstral-2512",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Mistral: Devstral 2 2512 is 9.1% cheaper on the balanced blend",
      "summary": "$0.88/M on 2026-08-30, $0.8/M on 2026-09-02 — a 9.1% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: mistralai/devstral-2512.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "mistralai/devstral-2512",
      "evidenceUrl": "/models/mistralai__devstral-2512/"
    },
    {
      "id": "price-change-2026-09-02-nvidia-nemotron-3-ultra-550b-a55b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "NVIDIA: Nemotron 3 Ultra is 7.4% cheaper on the balanced blend",
      "summary": "$1.35/M on 2026-08-30, $1.25/M on 2026-09-02 — a 7.4% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: nvidia/nemotron-3-ultra-550b-a55b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "nvidia/nemotron-3-ultra-550b-a55b",
      "evidenceUrl": "/models/nvidia__nemotron-3-ultra-550b-a55b/"
    },
    {
      "id": "price-change-2026-09-02-qwen-qwen3-235b-a22b-2507",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Qwen: Qwen3 235B A22B Instruct 2507 is 25.3% cheaper on the balanced blend",
      "summary": "$0.205/M on 2026-08-30, $0.1531/M on 2026-09-02 — a 25.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-235b-a22b-2507.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-235b-a22b-2507",
      "evidenceUrl": "/models/qwen__qwen3-235b-a22b-2507/"
    },
    {
      "id": "price-change-2026-09-02-qwen-qwen3-5-122b-a10b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Qwen: Qwen3.5-122B-A10B is 14.3% dearer on the balanced blend",
      "summary": "$0.715/M on 2026-08-30, $0.8175/M on 2026-09-02 — a 14.3% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.5-122b-a10b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.5-122b-a10b",
      "evidenceUrl": "/models/qwen__qwen3.5-122b-a10b/"
    },
    {
      "id": "price-change-2026-09-02-qwen-qwen3-6-27b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Qwen: Qwen3.6 27B is 29.8% dearer on the balanced blend",
      "summary": "$1.04/M on 2026-08-30, $1.35/M on 2026-09-02 — a 29.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.6-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.6-27b",
      "evidenceUrl": "/models/qwen__qwen3.6-27b/"
    },
    {
      "id": "price-change-2026-09-02-qwen-qwen3-6-35b-a3b",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Qwen: Qwen3.6 35B A3B is 15.5% cheaper on the balanced blend",
      "summary": "$0.355/M on 2026-08-30, $0.3/M on 2026-09-02 — a 15.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.6-35b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.6-35b-a3b",
      "evidenceUrl": "/models/qwen__qwen3.6-35b-a3b/"
    },
    {
      "id": "price-change-2026-09-02-qwen-qwen3-vl-30b-a3b-instruct",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Qwen: Qwen3 VL 30B A3B Instruct is 15.4% dearer on the balanced blend",
      "summary": "$0.2275/M on 2026-08-30, $0.2625/M on 2026-09-02 — a 15.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-vl-30b-a3b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-vl-30b-a3b-instruct",
      "evidenceUrl": "/models/qwen__qwen3-vl-30b-a3b-instruct/"
    },
    {
      "id": "price-change-2026-09-02-tencent-hy3",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Tencent: Hy3 is 60% dearer on the balanced blend",
      "summary": "$0.1444/M on 2026-08-30, $0.231/M on 2026-09-02 — a 60% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: tencent/hy3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-09-02-z-ai-glm-4-6",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Z.ai: GLM 4.6 is 13.1% cheaper on the balanced blend",
      "summary": "$0.875/M on 2026-08-30, $0.76/M on 2026-09-02 — a 13.1% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-4.6.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-4.6",
      "evidenceUrl": "/models/z-ai__glm-4.6/"
    },
    {
      "id": "price-change-2026-09-02-z-ai-glm-5-1",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Z.ai: GLM 5.1 is 23.3% cheaper on the balanced blend",
      "summary": "$1.94/M on 2026-08-30, $1.48/M on 2026-09-02 — a 23.3% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.1",
      "evidenceUrl": "/models/z-ai__glm-5.1/"
    },
    {
      "id": "price-change-2026-09-02-z-ai-glm-5-2",
      "date": "2026-09-02",
      "type": "price-change",
      "title": "Z.ai: GLM 5.2 is 18.8% cheaper on the balanced blend",
      "summary": "$1.83/M on 2026-08-30, $1.48/M on 2026-09-02 — a 18.8% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.2.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-30 → 2026-09-02. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "price-change-2026-08-26-deepseek-deepseek-v4-flash",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0423 is 65.8% dearer on the balanced blend",
      "summary": "$0.0611/M on 2026-08-25, $0.1013/M on 2026-08-26 — a 65.8% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash/"
    },
    {
      "id": "price-change-2026-08-26-deepseek-deepseek-v4-flash-0731",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Flash 0731 is 28.6% cheaper on the balanced blend",
      "summary": "$0.105/M on 2026-08-25, $0.075/M on 2026-08-26 — a 28.6% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-flash-0731.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-flash-0731",
      "evidenceUrl": "/models/deepseek__deepseek-v4-flash-0731/"
    },
    {
      "id": "price-change-2026-08-26-deepseek-deepseek-v4-pro",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0423 is 119.2% dearer on the balanced blend",
      "summary": "$0.4961/M on 2026-08-25, $1.09/M on 2026-08-26 — a 119.2% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro/"
    },
    {
      "id": "price-change-2026-08-26-deepseek-deepseek-v4-pro-0813",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "DeepSeek: DeepSeek V4 Pro 0813 is 17.6% dearer on the balanced blend",
      "summary": "$1.68/M on 2026-08-25, $1.98/M on 2026-08-26 — a 17.6% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: deepseek/deepseek-v4-pro-0813.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "deepseek/deepseek-v4-pro-0813",
      "evidenceUrl": "/models/deepseek__deepseek-v4-pro-0813/"
    },
    {
      "id": "price-change-2026-08-26-meta-llama-llama-3-3-70b-instruct",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Meta: Llama 3.3 70B Instruct is 358.1% dearer on the balanced blend",
      "summary": "$0.155/M on 2026-08-25, $0.71/M on 2026-08-26 — a 358.1% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-3.3-70b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-3.3-70b-instruct",
      "evidenceUrl": "/models/meta-llama__llama-3.3-70b-instruct/"
    },
    {
      "id": "price-change-2026-08-26-meta-llama-llama-4-scout",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Meta: Llama 4 Scout is 11.7% dearer on the balanced blend",
      "summary": "$0.15/M on 2026-08-25, $0.1675/M on 2026-08-26 — a 11.7% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: meta-llama/llama-4-scout.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "meta-llama/llama-4-scout",
      "evidenceUrl": "/models/meta-llama__llama-4-scout/"
    },
    {
      "id": "price-change-2026-08-26-minimax-minimax-m2-7",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "MiniMax: MiniMax M2.7 is 25% dearer on the balanced blend",
      "summary": "$0.42/M on 2026-08-25, $0.525/M on 2026-08-26 — a 25% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: minimax/minimax-m2.7.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "minimax/minimax-m2.7",
      "evidenceUrl": "/models/minimax__minimax-m2.7/"
    },
    {
      "id": "price-change-2026-08-26-moonshotai-kimi-k2-6",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "MoonshotAI: Kimi K2.6 is 75.4% dearer on the balanced blend",
      "summary": "$0.9761/M on 2026-08-25, $1.71/M on 2026-08-26 — a 75.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: moonshotai/kimi-k2.6.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "moonshotai/kimi-k2.6",
      "evidenceUrl": "/models/moonshotai__kimi-k2.6/"
    },
    {
      "id": "price-change-2026-08-26-qwen-qwen2-5-vl-72b-instruct",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Qwen: Qwen2.5 VL 72B Instruct is 55.9% cheaper on the balanced blend",
      "summary": "$0.85/M on 2026-08-25, $0.375/M on 2026-08-26 — a 55.9% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen2.5-vl-72b-instruct.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen2.5-vl-72b-instruct",
      "evidenceUrl": "/models/qwen__qwen2.5-vl-72b-instruct/"
    },
    {
      "id": "price-change-2026-08-26-qwen-qwen3-30b-a3b",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Qwen: Qwen3 30B A3B is 5.5% cheaper on the balanced blend",
      "summary": "$0.2275/M on 2026-08-25, $0.215/M on 2026-08-26 — a 5.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3-30b-a3b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3-30b-a3b",
      "evidenceUrl": "/models/qwen__qwen3-30b-a3b/"
    },
    {
      "id": "price-change-2026-08-26-qwen-qwen3-6-27b",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Qwen: Qwen3.6 27B is 23% cheaper on the balanced blend",
      "summary": "$1.35/M on 2026-08-25, $1.04/M on 2026-08-26 — a 23% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.6-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.6-27b",
      "evidenceUrl": "/models/qwen__qwen3.6-27b/"
    },
    {
      "id": "price-change-2026-08-26-qwen-qwen3-8-27b",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Qwen: Qwen3.8 27B is 8.9% cheaper on the balanced blend",
      "summary": "$1.05/M on 2026-08-25, $0.9562/M on 2026-08-26 — a 8.9% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: qwen/qwen3.8-27b.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "qwen/qwen3.8-27b",
      "evidenceUrl": "/models/qwen__qwen3.8-27b/"
    },
    {
      "id": "price-change-2026-08-26-tencent-hy3",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Tencent: Hy3 is 37.5% cheaper on the balanced blend",
      "summary": "$0.231/M on 2026-08-25, $0.1444/M on 2026-08-26 — a 37.5% fall in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: tencent/hy3.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "tencent/hy3",
      "evidenceUrl": "/models/tencent__hy3/"
    },
    {
      "id": "price-change-2026-08-26-z-ai-glm-5-1",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Z.ai: GLM 5.1 is 30.4% dearer on the balanced blend",
      "summary": "$1.48/M on 2026-08-25, $1.94/M on 2026-08-26 — a 30.4% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.1.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.1",
      "evidenceUrl": "/models/z-ai__glm-5.1/"
    },
    {
      "id": "price-change-2026-08-26-z-ai-glm-5-2",
      "date": "2026-08-26",
      "type": "price-change",
      "title": "Z.ai: GLM 5.2 is 23.2% dearer on the balanced blend",
      "summary": "$1.48/M on 2026-08-25, $1.83/M on 2026-08-26 — a 23.2% rise in the blended figure. The blend weights input and output, so the list prices behind it may have moved by different amounts and one of them may not have moved at all. Model: z-ai/glm-5.2.",
      "conditions": "Balanced-workload blend, effective price per million tokens, standard delivery. Rank snapshots 2026-08-25 → 2026-08-26. Threshold: 5% or more, the same one watch-upstream routes to a human.",
      "modelSlug": "z-ai/glm-5.2",
      "evidenceUrl": "/models/z-ai__glm-5.2/"
    },
    {
      "id": "correction-2026-08-25-22-the-delayed-analytics-beacon-used-the-wrong-collection-origin",
      "date": "2026-08-25",
      "type": "correction",
      "title": "Correction: the delayed analytics beacon used the wrong collection origin",
      "summary": "Logged 2026-08-25 as §22 of docs/EMPIRICAL-GROUNDING.md. The wire carries the fact that a correction was published; the log carries what was claimed, what is true, and how it was found. Corrections are not summarised here — shortening one is how a load-bearing qualification gets dropped. A vendor changing a price is not us being wrong; this is.",
      "conditions": "Correction log, docs/EMPIRICAL-GROUNDING.md §22, logged 2026-08-25.",
      "evidenceUrl": "/corrections/"
    },
    {
      "id": "correction-2026-08-24-17-the-tier-ladder-is-not-one-rung",
      "date": "2026-08-24",
      "type": "correction",
      "title": "Correction: the tier ladder is not one rung",
      "summary": "Logged 2026-08-24 as §17 of docs/EMPIRICAL-GROUNDING.md. The wire carries the fact that a correction was published; the log carries what was claimed, what is true, and how it was found. Corrections are not summarised here — shortening one is how a load-bearing qualification gets dropped. A vendor changing a price is not us being wrong; this is.",
      "conditions": "Correction log, docs/EMPIRICAL-GROUNDING.md §17, logged 2026-08-24.",
      "evidenceUrl": "/corrections/"
    },
    {
      "id": "correction-2026-08-24-18-r-0-409-was-a-small-sample-artefact-it-is-0-806",
      "date": "2026-08-24",
      "type": "correction",
      "title": "Correction: r = 0.409 was a small-sample artefact; it is 0.806",
      "summary": "Logged 2026-08-24 as §18 of docs/EMPIRICAL-GROUNDING.md. The wire carries the fact that a correction was published; the log carries what was claimed, what is true, and how it was found. Corrections are not summarised here — shortening one is how a load-bearing qualification gets dropped. A vendor changing a price is not us being wrong; this is.",
      "conditions": "Correction log, docs/EMPIRICAL-GROUNDING.md §18, logged 2026-08-24.",
      "evidenceUrl": "/corrections/"
    },
    {
      "id": "correction-2026-08-24-19-let-cloudflare-compress-it-made-every-asset-larger",
      "date": "2026-08-24",
      "type": "correction",
      "title": "Correction: \"let Cloudflare compress it\" made every asset larger",
      "summary": "Logged 2026-08-24 as §19 of docs/EMPIRICAL-GROUNDING.md. The wire carries the fact that a correction was published; the log carries what was claimed, what is true, and how it was found. Corrections are not summarised here — shortening one is how a load-bearing qualification gets dropped. A vendor changing a price is not us being wrong; this is.",
      "conditions": "Correction log, docs/EMPIRICAL-GROUNDING.md §19, logged 2026-08-24.",
      "evidenceUrl": "/corrections/"
    }
  ]
}