Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresDAXBench v2 (Maxim Anatsko with SQLBI) asks models to write DAX, Power BI's formula language, for 94 tasks on prepared Power BI models of the Contoso sample data: writing measures, calculated columns and functions, and fixing or optimizing broken ones, across basic, intermediate and advanced levels. Each answer runs inside real DAX probe queries in several report contexts, and a task counts as correct only if every probe and every required structural check passes. This is the share of the 94 tasks answered fully correctly. Version 1 used 30 different tasks and another scoring path and is not comparable.
649th strongest of 798 fitted benchmarks
Signal-to-noise +3.02
508th hardest of 798 fitted benchmarks
Capability needed 137.5
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | GPT-6 Astra | not reported | 96.8 | |
| 2 | Claude Opus 5.5 | not reported | 95.7 | |
| 3 | Gemini 4 Carbon | not reported | 93.6 | |
| 4 | GPT-6.1 Terra | not reported | 93.6 | |
| 5 | GPT-5.6 Sol | medium | 92.6 | |
| 6 | GPT-6.1 Sol | not reported | 92.6 | |
| 7 | Grok 4.7 | not reported | 92.6 | |
| 8 | Claude Fable 5.1 | not reported | 91.5 | |
| 9 | Claude Fable 5.2 | not reported | 88.3 | |
| 10 | Gemini 4 Argon | not reported | 88.3 | |
| 11 | GPT-5.5 | not reported | 88.3 | |
| 12 | Gemini 3.8 Flash | not reported | 87.2 | |
| 13 | Hy4 | not reported | 87.2 | |
| 14 | Claude Opus 5 | not reported | 86.2 | |
| 15 | Claude Sonnet 5.5 | not reported | 86.2 | |
| 16 | Grok 4.6 | not reported | 86.2 | |
| 17 | Kimi K3.1 Pro | — | 86.2 | |
| 18 | GPT-6 Luna | not reported | 85.1 | |
| 19 | Muse Spark 1.3 | not reported | 85.1 | |
| 20 | Muse Wonder | not reported | 84.0 | |
| 21 | Gemini 3.1 Pro Preview | not reported | 83.0 | |
| 22 | Qwen3.8 Max (0902) | not reported | 83.0 | |
| 23 | Claude Haiku 5.5 | not reported | 81.9 | |
| 24 | Kimi K2.5 | not reported | 81.9 | |
| 25 | MiMo-V2.6-Pro | not reported | 81.9 | |
| 26 | GLM-5.3 | not reported | 79.8 | |
| 27 | Kimi K2.7 Code | always on | 79.8 | |
| 28 | Kimi K3 | — | 79.8 | |
| 29 | GLM-5.3 Flash | not reported | 78.7 | |
| 30 | GPT-5.4 Mini | medium | 78.7 | |
| 31 | DeepSeek V4.1 Flash 1017 | not reported | 77.7 | |
| 32 | DeepSeek V4.1 Flash | not reported | 75.5 | |
| 33 | Ling 3.1 Flash | not reported | 74.5 | |
| 34 | Gemini 2.5 Pro | — | 73.4 | |
| 35 | GLM-4.5 | not reported | 73.4 | |
| 36 | Sonnet 5 | not reported | 72.3 | |
| 37 | GPT-5.1 | medium | 72.3 | |
| 38 | DeepSeek V4 Pro 0813 | not reported | 69.1 | |
| 39 | GLM-5 | not reported | 69.1 | |
| 40 | Gemini 3 Flash Preview | not reported | 67.0 | |
| 41 | MiniMax M3 | — | 67.0 | |
| 42 | GPT-5 Mini | — | 66.0 | |
| 43 | Mistral Large 4 Preview | not reported | 66.0 | |
| 44 | MiniMax M3.1 Flash | — | 63.8 | |
| 45 | Muse Glimmer 30B | not reported | 63.8 | |
| 46 | Opus 4.5 | — | 62.8 | |
| 47 | MiMo-V2.6-Flash | not reported | 62.8 | |
| 48 | Muse Ember | not reported | 62.8 | |
| 49 | Qwen3.6 27B | not reported | 62.8 | |
| 50 | Sonnet 4 | not reported | 61.7 | |
| 51 | DeepSeek-R1 | always on | 61.7 | |
| 52 | Kimi K2 Instruct | not supported | 59.6 | |
| 53 | Sonnet 4.5 | not reported | 58.5 | |
| 54 | gpt-oss-120b | not reported | 58.5 | |
| 55 | Ripple-65B Preview | unknown | 58.5 | |
| 56 | Gemma 4 26B A4B IT | not reported | 57.4 | |
| 57 | Qwen3.5-27B | not reported | 57.4 | |
| 58 | DeepSeek-V3.2 | not reported | 54.3 | |
| 59 | Gemini 2.5 Flash | — | 54.3 | |
| 60 | Mistral Large 2 (Instruct 2407) | not supported | 54.3 | |
| 61 | DeepSeek-V4-Flash-0731 | not reported | 53.2 | |
| 62 | Nemotron 3 Super 120B A12B | — | 53.2 | |
| 63 | Qwen3.8 27B | x-high | 52.1 | |
| 64 | MiniMax M2 | always on | 51.1 | |
| 65 | MiniMax M2.5 | always on | 51.1 | |
| 66 | GPT-4o (2024-11-20) | not supported | 50.0 | |
| 67 | Nemotron 3.5 Lightning 30B A3B (precision unspecified) | not reported | 47.9 | |
| 68 | Llama 4 Maverick Instruct | not supported | 46.8 | |
| 69 | Haiku 4.5 | not reported | 41.5 | |
| 70 | GPT-4o Mini | not supported | 40.4 | |
| 71 | Laguna S 2.1 | not reported | 40.4 | |
| 72 | Qwen3-30B-A3B | not reported | 40.4 | |
| 73 | gpt-oss-20b | not reported | 39.4 | |
| 74 | LFM2.5-24B-A2B | always on | 39.4 | |
| 75 | GLM 4.7 Flash | not reported | 38.3 | |
| 76 | Qwen2.5 Coder 32B Instruct | not supported | 36.2 | |
| 77 | Llama 3.3 70B Instruct | not supported | 33.0 | |
| 78 | Laguna XS 2.1 | not reported | 31.9 | |
| 79 | Gemma 3 27B IT | not supported | 30.9 | |
| 80 | Ministral 3 14B Instruct 2512 | not supported | 26.6 | |
| 81 | Llama 4 Scout Instruct | not supported | 25.5 | |
| 82 | MiniMax M1 | always on | 1.1 |
Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).