Atlas

Benchmarks

DAXBench v2

DAXBench 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.

Models
82
Top score
96.8
Leader
GPT-6 Astra
Updated
—
Resolution
~9 mo
Coarse
Saturation
Feb 2026
Reached
Signal
19thpercentile
195 benchmarks from 0.8 to 3.5224 benchmarks from 3.5 to 6.1133 benchmarks from 6.1 to 8.898 benchmarks from 8.8 to 11.566 benchmarks from 11.5 to 14.126 benchmarks from 14.1 to 16.815 benchmarks from 16.8 to 19.59 benchmarks from 19.5 to 22.18 benchmarks from 22.1 to 24.84 benchmarks from 24.8 to 27.55 benchmarks from 27.5 to 30.12 benchmarks from 30.1 to 32.84 benchmarks from 32.8 to 35.45 benchmarks from 35.4 to 38.11 benchmark from 38.1 to 40.82 benchmarks from 43.4 to 46.11 benchmark from 62.1 to 64.8

649th strongest of 798

Signal-to-noise +3.02

Difficulty
36thpercentile
1 benchmark from 5.5 to 14.41 benchmark from 14.4 to 23.22 benchmarks from 32.1 to 41.01 benchmark from 41.0 to 49.82 benchmarks from 49.8 to 58.79 benchmarks from 58.7 to 67.67 benchmarks from 67.6 to 76.418 benchmarks from 76.4 to 85.324 benchmarks from 85.3 to 94.241 benchmarks from 94.2 to 103.031 benchmarks from 103.0 to 111.939 benchmarks from 111.9 to 120.850 benchmarks from 120.8 to 129.676 benchmarks from 129.6 to 138.560 benchmarks from 138.5 to 147.4118 benchmarks from 147.4 to 156.2124 benchmarks from 156.2 to 165.1120 benchmarks from 165.1 to 174.039 benchmarks from 174.0 to 182.820 benchmarks from 182.8 to 191.710 benchmarks from 191.7 to 200.61 benchmark from 200.6 to 209.44 benchmarks from 209.4 to 218.3

508th hardest of 798

Capability needed 137.5

Organizations
82 dated models · 9 frontier modelsClosed weightsOpen weights
Data: DAXBenchatlas.village.dev

DAXBench v2 leaderboard

#ModelSettingScoreSource
1GPT-6 Astranot reported
96.8
2Claude Opus 5.5not reported
95.7
3Gemini 4 Carbonnot reported
93.6
4GPT-6.1 Terranot reported
93.6
5GPT-5.6 Solmedium
92.6
6GPT-6.1 Solnot reported
92.6
7Grok 4.7not reported
92.6
8Claude Fable 5.1not reported
91.5
9Claude Fable 5.2not reported
88.3
10Gemini 4 Argonnot reported
88.3
11GPT-5.5not reported
88.3
12Gemini 3.8 Flashnot reported
87.2
13Hy4not reported
87.2
14Claude Opus 5not reported
86.2
15Claude Sonnet 5.5not reported
86.2
16Grok 4.6not reported
86.2
17Kimi K3.1 Pro—
86.2
18GPT-6 Lunanot reported
85.1
19Muse Spark 1.3not reported
85.1
20Muse Wondernot reported
84.0
21Gemini 3.1 Pro Previewnot reported
83.0
22Qwen3.8 Max (0902)not reported
83.0
23Claude Haiku 5.5not reported
81.9
24Kimi K2.5not reported
81.9
25MiMo-V2.6-Pronot reported
81.9
26GLM-5.3not reported
79.8
27Kimi K2.7 Codealways on
79.8
28Kimi K3—
79.8
29GLM-5.3 Flashnot reported
78.7
30GPT-5.4 Minimedium
78.7
31DeepSeek V4.1 Flash 1017not reported
77.7
32DeepSeek V4.1 Flashnot reported
75.5
33Ling 3.1 Flashnot reported
74.5
34Gemini 2.5 Pro—
73.4
35GLM-4.5not reported
73.4
36Sonnet 5not reported
72.3
37GPT-5.1medium
72.3
38DeepSeek V4 Pro 0813not reported
69.1
39GLM-5not reported
69.1
40Gemini 3 Flash Previewnot reported
67.0
41MiniMax M3—
67.0
42GPT-5 Mini—
66.0
43Mistral Large 4 Previewnot reported
66.0
44MiniMax M3.1 Flash—
63.8
45Muse Glimmer 30Bnot reported
63.8
46Opus 4.5—
62.8
47MiMo-V2.6-Flashnot reported
62.8
48Muse Embernot reported
62.8
49Qwen3.6 27Bnot reported
62.8
50Sonnet 4not reported
61.7
51DeepSeek-R1always on
61.7
52Kimi K2 Instructnot supported
59.6
53Sonnet 4.5not reported
58.5
54gpt-oss-120bnot reported
58.5
55Ripple-65B Previewunknown
58.5
56Gemma 4 26B A4B ITnot reported
57.4
57Qwen3.5-27Bnot reported
57.4
58DeepSeek-V3.2not reported
54.3
59Gemini 2.5 Flash—
54.3
60Mistral Large 2 (Instruct 2407)not supported
54.3
61DeepSeek-V4-Flash-0731not reported
53.2
62Nemotron 3 Super 120B A12B—
53.2
63Qwen3.8 27Bx-high
52.1
64MiniMax M2always on
51.1
65MiniMax M2.5always on
51.1
66GPT-4o (2024-11-20)not supported
50.0
67Nemotron 3.5 Lightning 30B A3B (precision unspecified)not reported
47.9
68Llama 4 Maverick Instructnot supported
46.8
69Haiku 4.5not reported
41.5
70GPT-4o Mininot supported
40.4
71Laguna S 2.1not reported
40.4
72Qwen3-30B-A3Bnot reported
40.4
73gpt-oss-20bnot reported
39.4
74LFM2.5-24B-A2Balways on
39.4
75GLM 4.7 Flashnot reported
38.3
76Qwen2.5 Coder 32B Instructnot supported
36.2
77Llama 3.3 70B Instructnot supported
33.0
78Laguna XS 2.1not reported
31.9
79Gemma 3 27B ITnot supported
30.9
80Ministral 3 14B Instruct 2512not supported
26.6
81Llama 4 Scout Instructnot supported
25.5
82MiniMax M1always on
1.1

Related benchmarks

Most correlated

Benchmarks outside its family whose model rankings track it most reliably (Spearman ρ over at least 10 shared models).

Loading Atlas data…