Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresThe points score DAXBench v2 ranks by: each fully correct task earns its complexity weight (basic 1.0, intermediate 1.3, advanced 1.6) plus a style bonus for best practices, performance and clarity, and the total is a percentage of the 147.36 available points; incorrect tasks earn nothing. Diagnostic, because the weights and the style bonus make it a composite rather than an accuracy; the primary daxbench-v2 counts fully correct tasks.
No signal or difficulty estimate. Held out of the fit by design.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | GPT-6 Astra | not reported | 95.9 | |
| 2 | Claude Opus 5.5 | not reported | 95.4 | |
| 3 | Gemini 4 Carbon | not reported | 92.8 | |
| 4 | GPT-6.1 Terra | not reported | 91.9 | |
| 5 | GPT-5.6 Sol | medium | 91.5 | |
| 6 | GPT-6.1 Sol | not reported | 91.4 | |
| 7 | Grok 4.7 | not reported | 91.2 | |
| 8 | Claude Fable 5.1 | not reported | 90.7 | |
| 9 | Claude Fable 5.2 | not reported | 87.8 | |
| 10 | Gemini 4 Argon | not reported | 87.5 | |
| 11 | GPT-5.5 | not reported | 86.4 | |
| 12 | Hy4 | not reported | 85.7 | |
| 13 | Gemini 3.8 Flash | not reported | 85.6 | |
| 14 | Claude Opus 5 | not reported | 85.4 | |
| 15 | Claude Sonnet 5.5 | not reported | 84.6 | |
| 16 | Grok 4.6 | not reported | 83.9 | |
| 17 | Kimi K3.1 Pro | — | 83.9 | |
| 18 | GPT-6 Luna | not reported | 83.0 | |
| 19 | Muse Spark 1.3 | not reported | 82.6 | |
| 20 | Muse Wonder | not reported | 82.2 | |
| 21 | Qwen3.8 Max (0902) | not reported | 81.7 | |
| 22 | Gemini 3.1 Pro Preview | not reported | 81.2 | |
| 23 | MiMo-V2.6-Pro | not reported | 80.7 | |
| 24 | Kimi K2.5 | not reported | 79.7 | |
| 25 | Claude Haiku 5.5 | not reported | 79.4 | |
| 26 | Kimi K3 | — | 78.0 | |
| 27 | Kimi K2.7 Code | always on | 77.8 | |
| 28 | GLM-5.3 | not reported | 77.3 | |
| 29 | GLM-5.3 Flash | not reported | 75.8 | |
| 30 | GPT-5.4 Mini | medium | 75.8 | |
| 31 | DeepSeek V4.1 Flash 1017 | not reported | 74.5 | |
| 32 | DeepSeek V4.1 Flash | not reported | 72.8 | |
| 33 | Ling 3.1 Flash | not reported | 72.7 | |
| 34 | Gemini 2.5 Pro | — | 71.9 | |
| 35 | GLM-4.5 | not reported | 71.0 | |
| 36 | Sonnet 5 | not reported | 70.4 | |
| 37 | GPT-5.1 | medium | 70.4 | |
| 38 | DeepSeek V4 Pro 0813 | not reported | 66.3 | |
| 39 | GLM-5 | not reported | 66.0 | |
| 40 | MiniMax M3 | — | 64.2 | |
| 41 | Gemini 3 Flash Preview | not reported | 63.5 | |
| 42 | Mistral Large 4 Preview | not reported | 63.2 | |
| 43 | GPT-5 Mini | — | 63.0 | |
| 44 | MiniMax M3.1 Flash | — | 61.0 | |
| 45 | Muse Glimmer 30B | not reported | 60.0 | |
| 46 | Opus 4.5 | — | 59.6 | |
| 47 | Muse Ember | not reported | 59.4 | |
| 48 | Qwen3.6 27B | not reported | 59.0 | |
| 49 | MiMo-V2.6-Flash | not reported | 58.9 | |
| 50 | Sonnet 4 | not reported | 58.4 | |
| 51 | DeepSeek-R1 | always on | 57.6 | |
| 52 | Kimi K2 Instruct | not supported | 57.1 | |
| 53 | Sonnet 4.5 | not reported | 55.3 | |
| 54 | Ripple-65B Preview | unknown | 55.0 | |
| 55 | gpt-oss-120b | not reported | 55.0 | |
| 56 | Qwen3.5-27B | not reported | 53.8 | |
| 57 | Gemma 4 26B A4B IT | not reported | 53.2 | |
| 58 | Mistral Large 2 (Instruct 2407) | not supported | 49.9 | |
| 59 | Gemini 2.5 Flash | — | 49.9 | |
| 60 | DeepSeek-V3.2 | not reported | 49.8 | |
| 61 | DeepSeek-V4-Flash-0731 | not reported | 49.2 | |
| 62 | Nemotron 3 Super 120B A12B | — | 48.5 | |
| 63 | Qwen3.8 27B | x-high | 48.2 | |
| 64 | MiniMax M2 | always on | 47.9 | |
| 65 | MiniMax M2.5 | always on | 46.7 | |
| 66 | GPT-4o (2024-11-20) | not supported | 45.2 | |
| 67 | Nemotron 3.5 Lightning 30B A3B (precision unspecified) | not reported | 44.0 | |
| 68 | Llama 4 Maverick Instruct | not supported | 42.0 | |
| 69 | Laguna S 2.1 | not reported | 37.6 | |
| 70 | Qwen3-30B-A3B | not reported | 37.0 | |
| 71 | Haiku 4.5 | not reported | 37.0 | |
| 72 | GPT-4o Mini | not supported | 35.7 | |
| 73 | LFM2.5-24B-A2B | always on | 35.3 | |
| 74 | gpt-oss-20b | not reported | 34.4 | |
| 75 | GLM 4.7 Flash | not reported | 33.9 | |
| 76 | Qwen2.5 Coder 32B Instruct | not supported | 31.9 | |
| 77 | Llama 3.3 70B Instruct | not supported | 28.3 | |
| 78 | Laguna XS 2.1 | not reported | 27.5 | |
| 79 | Gemma 3 27B IT | not supported | 26.6 | |
| 80 | Ministral 3 14B Instruct 2512 | not supported | 23.2 | |
| 81 | Llama 4 Scout Instruct | not supported | 21.4 | |
| 82 | MiniMax M1 | always on | 0.8 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).