Is Llama 3.1 70B Instruct a good deal?

Whether anything in this catalogue beats Llama 3.1 70B Instruct on both quality and price, and what you give up if it does. A computation on the current catalogue, not an opinion.

12 undominated of 144 · Nov 22, 2026

As of Nov 22, 2026, Llama 3.1 70B Instruct is dominated for Balanced on LMArena. MiMo-V2.6-Pro scores 231.8 higher and costs 25% less, with a covering envelope. 12 of 144 rated, priced standard models are undominated.

Inspect model evidence Compare differences & requirements

Current model

Llama 3.1 70B Instruct

Balanced

3 tokens in per 1 out

MiMo-V2.6-Pro is both better and cheaper than Llama 3.1 70B Instruct: 231.8 points higher on LMArena and 25% less per million tokens, $0.180 cheaper at this mix.

Llama 3.1 70B Instruct

LMArena Elo · higher is better

Scale starts at 1040 Elo

1260.9

Effective $/M · Balanced · lower is better

$0.720/M

MiMo-V2.6-Pro

LMArena Elo · higher is better

Scale starts at 1040 Elo

1492.7

Effective $/M · Balanced · lower is better

$0.540/M

Llama 3.1 70B Instruct takes text, returns up to 8,192 tokens from a 131,072-token context, and is listed by 1 seller.

Compared against 144 rated, priced models on this lens: 48 models dominate it and give up nothing, 4 more dominate it but give something up. Llama 3.1 70B Instruct scores 1260.9 at $0.720 per million tokens for this mix.

Envelope-safe replacements

Each row scores at least as high, costs no more, and covers this model’s context, output, modalities, tools, and reasoning. A cheaper narrower model is not listed here.

ModelLMArenaEffective $/MYou save
MiMo-V2.6-Pro1492.7 +231.8$0.540/M25%
GLM 5.3 Flash1471.4 +210.5$0.148/M79%
DeepSeek V4.1 Flash1463.0 +202.1$0.227/M68%
Qwen3.7 Plus1454.1 +193.2$0.560/M22%
MiMo-V2.6-Flash1449.8 +188.9$0.175/M76%
Gemma 4 31B1442.5 +181.6$0.205/M72%
Hy31439.0 +178.1$0.231/M68%
Gemma 4 26B A4B 1434.2 +173.3$0.123/M83%
DeepSeek V4 Flash 04231431.9 +171.0$0.113/M84%
GPT-5.6 Luna1430.9 +170.0$0.450/M38%
MiniMax M31426.5 +165.6$0.525/M27%
DeepSeek V3.21424.6 +163.7$0.394/M45%
Beam 1.0 501B A23B1424.3 +163.4$0.600/M17%
DeepSeek V3.2 Exp1422.3 +161.4$0.305/M58%
Qwen3 VL 235B A22B Instruct1420.0 +159.1$0.370/M49%
DeepSeek V3.11419.8 +158.9$0.453/M37%
Qwen3 235B A22B Instruct 25071419.2 +158.3$0.205/M72%
Qwen3.5-122B-A10B1417.5 +156.6$0.715/M1%
DeepSeek V3.1 Terminus1417.3 +156.4$0.453/M37%
Qwen3 Next 80B A3B Instruct1417.2 +156.3$0.343/M52%
Gemini 3.1 Flash Lite Preview1416.2 +155.3$0.563/M22%
Inkling Small1414.3 +153.4$0.638/M11%
Qwen3.5-27B1408.6 +147.7$0.536/M26%
Inkling 2 Small1402.6 +141.7$0.475/M34%
Claude Haiku 5.51397.5 +136.6$0.200/M72%
Qwen3.5-Flash1397.1 +136.2$0.114/M84%
Qwen3.5-35B-A3B1394.8 +133.9$0.355/M51%
GPT-6 Luna1390.2 +129.3$0.200/M72%
Solar Pro 41388.3 +127.4$0.525/M27%
Qwen3 30B A3B Instruct 25071383.7 +122.8$0.143/M80%
GLM 4.5 Air1383.7 +122.8$0.310/M57%
GLM 4.6V1376.4 +115.5$0.450/M38%
DeepSeek V3 03241374.6 +113.7$0.438/M39%
GPT-5 Mini1373.5 +112.6$0.688/M5%
GPT-5.4 Nano1371.8 +110.9$0.463/M36%
gpt-oss-120b1365.4 +104.5$0.060/M92%
MiniMax M2.51359.2 +98.3$0.440/M39%
Gemma 3 27B1357.8 +96.9$0.100/M86%
GLM 4.7 Flash1350.7 +89.8$0.145/MPrice under confirmation since Nov 21, 202680%
GPT-4.1 Mini1340.5 +79.6$0.700/M3%
Qwen3 32B1340.1 +79.2$0.248/M66%
Gemma 3 12B1334.2 +73.3$0.075/M90%
DeepSeek V31332.5 +71.6$0.500/M31%
GPT-5 Nano1320.0 +59.1$0.138/M81%
Granite 4.2 8B1318.8 +57.9$0.083/M89%
gpt-oss-20b1287.6 +26.7$0.034/M95%
GPT-4.1 Nano1284.8 +23.9$0.175/M76%
Llama 3.3 70B Instruct1274.3 +13.4$0.280/M61%

Higher score, lower price, named losses

Not a drop-in. The loss is why these are not a recommendation.

ModelLMArenaEffective $/MYou give up
Qwen3 30B A3B1316.8$0.215/Mcontext
Gemma 3 4B1290.8$0.043/Mtools
GPT-4o-mini (2024-07-18)1286.4$0.263/Mcontext
Qwen2.5 72B Instruct1269.1$0.370/Mcontext
MiMo-V2.6-Pro1492.7 · $0.540/MLlama 3.1 70B Instruct1260.9 · $0.720/MLMArena EloEffective $/M · Balanced
144 rated, priced standard models. Chartreuse is the frontier. Cobalt is the model you named, when it is not on the staircase.

Open the frontier

The link keeps the model and mix, using the current catalogue. Monthly spend and switching cost stay in this browser tab and are left out of the link.

Saved decision references

Save the model, workload, catalogue date and capability-preservation rule in this browser. Spend, switching cost and bill contents are not saved. No account or notifications.

Constraint: replacements must preserve the model’s capabilities; any losses remain named trade-offs.

Historical decisions cannot be fully replayed from saved references: past prices, scores and capability evidence are not stored.

Evidence & Ask