Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresShare of DAXBench v2's 94 answers that were syntactically valid DAX, whether or not they returned the expected results. A diagnostic beside the primary daxbench-v2, which counts fully correct tasks.
No signal or difficulty estimate. Held out of the fit by design.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Gemini 3.8 Flash | not reported | 100.0 | |
| 2 | Grok 4.7 | not reported | 100.0 | |
| 3 | Gemini 4 Argon | not reported | 98.9 | |
| 4 | GPT-6 Astra | not reported | 98.9 | |
| 5 | Kimi K2.7 Code | always on | 98.9 | |
| 6 | Claude Fable 5.1 | not reported | 97.9 | |
| 7 | Claude Haiku 5.5 | not reported | 97.9 | |
| 8 | Claude Opus 5 | not reported | 97.9 | |
| 9 | Claude Opus 5.5 | not reported | 97.9 | |
| 10 | Claude Sonnet 5.5 | not reported | 97.9 | |
| 11 | GPT-5.4 Mini | medium | 97.9 | |
| 12 | Kimi K2.5 | not reported | 97.9 | |
| 13 | Qwen3.8 Max (0902) | not reported | 97.9 | |
| 14 | Sonnet 5 | not reported | 96.8 | |
| 15 | Gemini 3.1 Pro Preview | not reported | 96.8 | |
| 16 | GPT-5.6 Sol | medium | 96.8 | |
| 17 | GPT-6.1 Sol | not reported | 96.8 | |
| 18 | GPT-6.1 Terra | not reported | 96.8 | |
| 19 | Gemini 4 Carbon | not reported | 95.7 | |
| 20 | GPT-5.5 | not reported | 95.7 | |
| 21 | Kimi K3 | — | 95.7 | |
| 22 | Muse Spark 1.3 | not reported | 95.7 | |
| 23 | Qwen3.6 27B | not reported | 95.7 | |
| 24 | GPT-5.1 | medium | 94.7 | |
| 25 | Sonnet 4 | not reported | 93.6 | |
| 26 | Gemini 2.5 Pro | — | 93.6 | |
| 27 | Gemini 3 Flash Preview | not reported | 93.6 | |
| 28 | GPT-6 Luna | not reported | 93.6 | |
| 29 | Grok 4.6 | not reported | 93.6 | |
| 30 | Kimi K3.1 Pro | — | 93.6 | |
| 31 | MiMo-V2.6-Flash | not reported | 93.6 | |
| 32 | DeepSeek-R1 | always on | 92.6 | |
| 33 | DeepSeek V4.1 Flash 1017 | not reported | 92.6 | |
| 34 | GLM-5.3 | not reported | 92.6 | |
| 35 | gpt-oss-120b | not reported | 92.6 | |
| 36 | Muse Wonder | not reported | 92.6 | |
| 37 | Opus 4.5 | — | 91.5 | |
| 38 | DeepSeek V4.1 Flash | not reported | 91.5 | |
| 39 | Gemini 2.5 Flash | — | 91.5 | |
| 40 | GLM-4.5 | not reported | 91.5 | |
| 41 | GLM-5.3 Flash | not reported | 91.5 | |
| 42 | Hy4 | not reported | 91.5 | |
| 43 | MiMo-V2.6-Pro | not reported | 91.5 | |
| 44 | Gemma 4 26B A4B IT | not reported | 90.4 | |
| 45 | GPT-4o (2024-11-20) | not supported | 90.4 | |
| 46 | GPT-5 Mini | — | 90.4 | |
| 47 | Ling 3.1 Flash | not reported | 90.4 | |
| 48 | MiniMax M3 | — | 90.4 | |
| 49 | Mistral Large 2 (Instruct 2407) | not supported | 90.4 | |
| 50 | Muse Glimmer 30B | not reported | 90.4 | |
| 51 | GLM-5 | not reported | 89.4 | |
| 52 | Qwen3.5-27B | not reported | 89.4 | |
| 53 | Claude Fable 5.2 | not reported | 88.3 | |
| 54 | Kimi K2 Instruct | not supported | 88.3 | |
| 55 | Sonnet 4.5 | not reported | 87.2 | |
| 56 | GPT-4o Mini | not supported | 87.2 | |
| 57 | Mistral Large 4 Preview | not reported | 87.2 | |
| 58 | Muse Ember | not reported | 87.2 | |
| 59 | DeepSeek V4 Pro 0813 | not reported | 86.2 | |
| 60 | Qwen3.8 27B | low | 85.1 | |
| 61 | gpt-oss-20b | not reported | 83.0 | |
| 62 | MiniMax M2.5 | always on | 83.0 | |
| 63 | DeepSeek-V3.2 | not reported | 81.9 | |
| 64 | Laguna S 2.1 | not reported | 80.9 | |
| 65 | MiniMax M3.1 Flash | — | 80.9 | |
| 66 | Qwen2.5 Coder 32B Instruct | not supported | 79.8 | |
| 67 | Ripple-65B Preview | unknown | 79.8 | |
| 68 | Nemotron 3 Super 120B A12B | — | 78.7 | |
| 69 | Haiku 4.5 | not reported | 77.7 | |
| 70 | Laguna XS 2.1 | not reported | 76.6 | |
| 71 | Llama 4 Maverick Instruct | not supported | 76.6 | |
| 72 | MiniMax M2 | always on | 71.3 | |
| 73 | Gemma 3 27B IT | not supported | 69.1 | |
| 74 | Llama 3.3 70B Instruct | not supported | 69.1 | |
| 75 | GLM 4.7 Flash | not reported | 68.1 | |
| 76 | Nemotron 3.5 Lightning 30B A3B (precision unspecified) | not reported | 67.0 | |
| 77 | Qwen3-30B-A3B | not reported | 67.0 | |
| 78 | Llama 4 Scout Instruct | not supported | 63.8 | |
| 79 | LFM2.5-24B-A2B | always on | 61.7 | |
| 80 | Ministral 3 14B Instruct 2512 | not supported | 59.6 | |
| 81 | DeepSeek-V4-Flash-0731 | not reported | 58.5 | |
| 82 | MiniMax M1 | always on | 6.4 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).