Mei —or, the Models Engineering Index— calculates language models' practical performance on engineering tasks.
Mei calculates language models' engineering performance.
Performance cost tokens .
30 40 50 60 70 4 Neon 69.5 Fable 5.2 65.8 Opus 5.5 63.1 6.2 Astra 60.5 4 Hydrogen 57.2 Sonnet 5.5 57.2 6.2 Sol 56.9 Wonder Full 49.3 K3.1 Pro 49.0 4-Max 49.0 6.1 Terra 47.2 Wonder 46.4 V4.1 Flash 44.1 Grok 4.7 42.7 Haiku 5.5 42.5 GLM 5.3 42.0 Hy4 37.3 V2.6-Pro 35.4 6 Luna 34.8
8× 4× 2× 1× 0.5× 0.25× 0.125× 0.0625× 6 Luna 0.065× V4.1 Flash 0.13× V2.6-Pro 0.13× Hy4 0.33× Haiku 5.5 0.4× 6.2 Sol 0.4× 6.1 Terra 0.83× Wonder Full 1.1× K3.1 Pro 1.2× 4 Hydrogen 1.2× Wonder 1.3× 6.2 Astra 1.4× 4 Neon 1.9× GLM 5.3 2.2× Sonnet 5.5 2.5× Opus 5.5 3.5× Grok 4.7 3.9× Fable 5.2 6.5× 4-Max 7.3×
2× 0.5× 6.2 Astra 0.29× 6.2 Sol 0.35× Wonder Full 0.49× 6 Luna 0.62× Wonder 0.63× K3.1 Pro 0.74× Grok 4.7 0.9× V2.6-Pro 0.93× 4 Neon 1× 4 Hydrogen 1.1× Hy4 1.1× 4-Max 1.2× V4.1 Flash 1.4× GLM 5.3 1.5× Haiku 5.5 1.6× Opus 5.5 1.7× 6.1 Terra 1.7× Fable 5.2 2.1× Sonnet 5.5 2.5×
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 Opus 5.5 6.2 Astra 4 Hydrogen Sonnet 5.5 6.2 Sol Wonder Full K3.1 Pro 4-Max 6.1 Terra Wonder V4.1 Flash Grok 4.7 Haiku 5.5 GLM 5.3 Hy4 V2.6-Pro 6 Luna
30 40 50 60 70 2× 0.5× 4 Neon Fable 5.2 Opus 5.5 6.2 Astra 4 Hydrogen Sonnet 5.5 6.2 Sol Wonder Full K3.1 Pro 4-Max 6.1 Terra Wonder V4.1 Flash Grok 4.7 Haiku 5.5 GLM 5.3 Hy4 V2.6-Pro 6 Luna
Rankings
Index bar
1 4 Neon Google 69.5 1.9× 1× 2 Fable 5.2 Anthropic 65.8 6.5× 2.1× 3 Opus 5.5 Anthropic 63.1 3.5× 1.7× 4 6.2 Astra OpenAI 60.5 1.4× 0.29× 5 4 Hydrogen Google 57.2 1.2× 1.1× 6 Sonnet 5.5 Anthropic 57.2 2.5× 2.5× 7 6.2 Sol OpenAI 56.9 0.4× 0.35× 8 Wonder Full Meta 49.3 1.1× 0.49× 9 K3.1 Pro Moonshot 49.0 1.2× 0.74× 10 4-Max Alibaba 49.0 7.3× 1.2× 11 6.1 Terra OpenAI 47.2 0.83× 1.7× 12 Wonder Meta 46.4 1.3× 0.63× 13 V4.1 Flash DeepSeek 44.1 0.13× 1.4× 14 Grok 4.7 xAI 42.7 3.9× 0.9× 15 Haiku 5.5 Anthropic 42.5 0.4× 1.6× 16 GLM 5.3 Z.ai 42.0 2.2× 1.5× 17 Hy4 Tencent 37.3 0.33× 1.1× 18 V2.6-Pro Xiaomi 35.4 0.13× 0.93× 19 6 Luna OpenAI 34.8 0.065× 0.62×
Comparing and
25 50 75 100 CursorBench FrontierCode Code Migration IOI ProgramBench Terminal-Bench Vibe Code
77.8 72.8 79.6 63.3 77.3 74.2 80.8 71.5 79.9