Mei —or, the Models Engineering Index— calculates language models' practical performance on engineering tasks.
Mei calculates language models' engineering performance.
Each model's score is a composite index across 14 coding benchmarks, adjusted for benchmark difficulty and missing results so models average 50. Cost and token use are adjusted the same way and shown relative to the typical model (1×).
Performance cost tokens .
30 40 50 60 70 4 Neon 68.7 Fable 5.2 67.2 6.2 Astra 67.1 Opus 5.5 61.9 6.2 Sol 59.8 Sonnet 5.5 59.6 4 Carbon 58.4 4 Hydrogen 55.1 4-Max 52.0 Wonder 50.1 6.1 Terra 49.7 Wonder Full 49.5 K3.1 Pro 49.4 Ember 46.9 V4.1 Flash 45.4 Haiku 5.5 44.6 Grok 4.7 43.2 GLM 5.3 42.2 Hy4 41.4 Inkling 2 Small 39.2 V2.6-Pro 39.1 6 Luna 37.1 M3.1 Flash 36.8 Large 4 35.5
8× 4× 2× 1× 0.5× 0.25× 0.125× 0.0625× Ember 0.044× 6 Luna 0.067× V4.1 Flash 0.15× V2.6-Pro 0.15× M3.1 Flash 0.26× Hy4 0.4× 6.2 Sol 0.46× Inkling 2 Small 0.54× Haiku 5.5 0.58× 6.1 Terra 1× Large 4 1× K3.1 Pro 1.4× 4 Hydrogen 1.5× 6.2 Astra 1.9× 4 Neon 2.1× Wonder 2.2× Wonder Full 2.2× 4 Carbon 2.3× GLM 5.3 2.4× Sonnet 5.5 3.7× Opus 5.5 4.5× Grok 4.7 4.6× Fable 5.2 6.2× 4-Max 9.5×
2× 0.5× 6.2 Sol 0.28× 6.2 Astra 0.38× 6 Luna 0.49× Ember 0.52× K3.1 Pro 0.66× Wonder Full 0.7× Grok 4.7 0.76× 4 Carbon 0.77× 4 Neon 0.8× Wonder 0.81× V2.6-Pro 0.86× Large 4 0.93× 4 Hydrogen 0.93× Hy4 1.1× 4-Max 1.2× GLM 5.3 1.3× V4.1 Flash 1.3× Fable 5.2 1.4× 6.1 Terra 1.6× Opus 5.5 1.6× Haiku 5.5 1.7× M3.1 Flash 2.4× Inkling 2 Small 2.8× Sonnet 5.5 3.2×
Performance versus
cost tokens .
30 40 50 60 70 8× 4× 2× 1× 0.5× 0.25× 0.125× 0.0625× 4 Neon Fable 5.2 6.2 Astra Opus 5.5 6.2 Sol Sonnet 5.5 4 Carbon 4 Hydrogen 4-Max Wonder 6.1 Terra Wonder Full K3.1 Pro Ember V4.1 Flash Haiku 5.5 Grok 4.7 GLM 5.3 Hy4 Inkling 2 Small V2.6-Pro 6 Luna M3.1 Flash Large 4
30 40 50 60 70 2× 0.5× 4 Neon Fable 5.2 6.2 Astra Opus 5.5 6.2 Sol Sonnet 5.5 4 Carbon 4 Hydrogen 4-Max Wonder 6.1 Terra Wonder Full K3.1 Pro Ember V4.1 Flash Haiku 5.5 Grok 4.7 GLM 5.3 Hy4 Inkling 2 Small V2.6-Pro 6 Luna M3.1 Flash Large 4
Help spread the word with a GitHub star (opens in a new tab)
Rankings
Index bar
1 4 Neon Google 68.7 2.1× 0.8× 2 Fable 5.2 Anthropic 67.2 6.2× 1.4× 3 6.2 Astra OpenAI 67.1 1.9× 0.38× 4 Opus 5.5 Anthropic 61.9 4.5× 1.6× 5 6.2 Sol OpenAI 59.8 0.46× 0.28× 6 Sonnet 5.5 Anthropic 59.6 3.7× 3.2× 7 4 Carbon Google 58.4 2.3× 0.77× 8 4 Hydrogen Google 55.1 1.5× 0.93× 9 4-Max Alibaba 52.0 9.5× 1.2× 10 Wonder Meta 50.1 2.2× 0.81× 11 6.1 Terra OpenAI 49.7 1× 1.6× 12 Wonder Full Meta 49.5 2.2× 0.7× 13 K3.1 Pro Moonshot 49.4 1.4× 0.66× 14 Ember Meta 46.9 0.044× 0.52× 15 V4.1 Flash DeepSeek 45.4 0.15× 1.3× 16 Haiku 5.5 Anthropic 44.6 0.58× 1.7× 17 Grok 4.7 xAI 43.2 4.6× 0.76× 18 GLM 5.3 Z.ai 42.2 2.4× 1.3× 19 Hy4 Tencent 41.4 0.4× 1.1× 20 Inkling 2 Small Thinking Machines 39.2 0.54× 2.8× 21 V2.6-Pro Xiaomi 39.1 0.15× 0.86× 22 6 Luna OpenAI 37.1 0.067× 0.49× 23 M3.1 Flash MiniMax 36.8 0.26× 2.4× 24 Large 4 Mistral 35.5 1× 0.93×