Atlas
Benchmarks
331 sources · 1,258 models × 2,980 benchmarks · 155,643 scoresMKA Regression WSR (Artificial Analysis)
The baseline-competence half of Artificial Analysis' MKA rubric: on the same 200 private tasks, the mean over tasks of the weighted share of the criteria the benchmark's authors label Regression, the easier checks that a deliverable meets basic requirements. It sits near the top of its range for strong models, so the Climb WSR headline separates them.
Models
19
Top score
96.8
Leader
Claude Sonnet 5.5
Updated
—
Resolution
—
Saturation
—
Signal
95thpercentile
37th strongest of 785 fitted benchmarks
Signal-to-noise +20.57 for the MKA (Artificial Analysis) family
Difficulty
Not estimated
Another benchmark represents this family in the fit.
Organizations
18 dated models · 4 frontier modelsClosed weightsOpen weights
Data: Artificial Analysisatlas.village.dev
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | max · Stirrup | 96.8 | |
| 2 | Gemini 4 Argon | high · Stirrup | 96.4 | |
| 3 | Gemini 4 Carbon | high · Stirrup | 96.4 | |
| 4 | Muse Wonder | max · Stirrup | 95.9 | |
| 5 | GPT-6 Astra | max · Stirrup | 95.2 | |
| 6 | GPT-6.1 Sol | max · Stirrup | 94.6 | |
| 7 | Kimi K3.1 Pro | — · Stirrup | 93.8 | |
| 8 | Gemini 3.8 Flash | high · Stirrup | 93.6 | |
| 9 | Claude Haiku 5.5 | max · Stirrup | 92.8 | |
| 10 | DeepSeek V4.1 Flash 1017 | max · Stirrup | 92.5 | |
| 11 | MiMo-V2.6-Pro | thinking · Stirrup | 91.7 | |
| 12 | Grok 4.7 | x-high · Stirrup | 91.7 | |
| 13 | GPT-6.1 Terra | max · Stirrup | 91.1 | |
| 14 | GLM-5.3 Flash | max · Stirrup | 90.9 | |
| 15 | Mistral Large 4 | high · Stirrup | 90.2 | |
| 16 | Qwen3.8 Max (0902) | thinking · Stirrup | 89.2 | |
| 17 | GPT-6 Luna | max · Stirrup | 89.1 | |
| 18 | MiniMax M3.1 Flash | — · Stirrup | 87.6 | |
| 19 | Step 5 Preview | — · Stirrup | 80.9 |
