价值前沿

所有模型,按价格与能力作图。那条阶梯线就是前沿:没有模型能在两个维度上同时胜过它们。位于其下方和右侧的一切都是严格更差的选择——总有既更聪明又更便宜的东西。

Changing the workload below moves the frontier less than you would expect: 91% of frontier models are on it under every workload (only Solar Pro 4 moves). What does change is the price — the same model swings up to across these shapes. So the control is not about which models to consider. It is about what you would actually pay for them.

10 未被压制(共 106 个) 96 个更差的选择
负载类型
测评项
$0.030$0.100$0.300$1.00$3.00$10.00102030405060每百万 tokens 的实际 $ — 对数刻度通用 能力
  1. 1 Ling-3.0-flashinclusionAI 37.8 $0.032/M 最便宜的一档
  2. 2 Solar Pro 4Upstage 41.6 $0.052/M +3.8,代价是 1.7× 的价格
  3. 3 DeepSeek V4 Flash 0423DeepSeek 42.1 $0.061/M +0.5,代价是 1.2× 的价格
  4. 4 DeepSeek V4 Flash 0731DeepSeek 51.8 $0.105/M +9.7,代价是 1.7× 的价格
  5. 5 GPT-5.6 LunaOpenAI 52.3 $0.450/M +0.5,代价是 4.3× 的价格
  6. 6 Gemini 3.7 FlashGoogle 56.0 $0.750/M +3.7,代价是 1.7× 的价格
  7. 7 Muse Spark 1.2Meta 56.8 $2.00/M +0.8,代价是 2.7× 的价格
  8. 8 GLM 5.3Z.ai 59.5 $2.15/M +2.7,代价是 1.1× 的价格
  9. 9 Grok 4.6xAI 60.9 $3.00/M +1.4,代价是 1.4× 的价格
  10. 10 Claude Opus 5Anthropic 63.1 $10.00/M +2.2,代价是 3.3× 的价格
Markdown for LLMs