Atlas
Benchmarks
326 sources · 1,256 models × 2,956 benchmarks · 154,484 scoresVIALS-AA (pass@3)
The share of tasks answered correctly at least once in three repeats. Artificial Analysis' own run of Handshake AI's 161-task VIALS life-sciences image set in Stirrup with Python and no network, plus an image-viewing tool for vision-capable models, three repeats per task and GPT-6 Luna (medium) grading. A repeat with no answer scores zero. Handshake's own protocol is the vials benchmark, so this stays separate.
Models
19
Top score
73.9
Leader
Kimi K3.1 Pro
Updated
—
Resolution
—
Saturation
—
Signal
42ndpercentile
450th strongest of 779 fitted benchmarks
Signal-to-noise +5.00 for the VIALS family
Difficulty
Not estimated
Another benchmark represents this family in the fit.
19 dated models · 4 frontier modelsClosed weightsOpen weights
Data: Artificial Analysisatlas.village.dev
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Kimi K3.1 Pro | — · Stirrup | 73.9 | |
| 2 | Muse Wonder | max · Stirrup | 73.9 | |
| 3 | GPT-6 Astra | max · Stirrup | 73.3 | |
| 4 | Gemini 4 Argon | high · Stirrup | 72.7 | |
| 5 | Qwen3.8 Max (0902) | thinking · Stirrup | 72.7 | |
| 6 | Inkling 2 Small | 0.99 · Stirrup | 72.0 | |
| 7 | Claude Haiku 5.5 | max · Stirrup | 71.4 | |
| 8 | GPT-6.1 Terra | max · Stirrup | 71.4 | |
| 9 | Claude Sonnet 5.5 | max · Stirrup | 70.8 | |
| 10 | Gemini 4 Carbon | high · Stirrup | 70.8 | |
| 11 | GPT-6.1 Sol | max · Stirrup | 70.8 | |
| 12 | MiMo-V2.6-Pro | thinking · Stirrup | 70.8 | |
| 13 | MiniMax M3.1 Flash | — · Stirrup | 70.8 | |
| 14 | GLM-5.3 Flash | max · Stirrup | 67.1 | |
| 15 | DeepSeek V4.1 Flash 1017 | max · Stirrup | 66.5 | |
| 16 | Grok 4.7 | x-high · Stirrup | 65.8 | |
| 17 | Gemini 3.8 Flash | high · Stirrup | 64.0 | |
| 18 | Mistral Large 4 | high · Stirrup | 62.7 | |
| 19 | GPT-6 Luna | max · Stirrup | 61.5 |
