Atlas
Benchmarks
331 sources · 1,258 models × 2,980 benchmarks · 155,643 scoresNeuroscienceBench V1 (91 evaluations)
Tests whether agents can reach correct scientific conclusions from neuroscience research data across sequencing, electrophysiology, optical physiology, EEG, brain-computer interface decoding, cell and tissue imaging, behavioral scoring and clinical biomarkers, in 91 evaluations. Scores are the share of graded runs that reach the correct conclusion: refusals, timeouts and harness errors are not graded and leave the denominator, so a configuration that refuses more is scored on fewer runs. Diagnostic for that reason; benchmarks.bio also excludes it from its overall Capabilities score.
Models
14
Top score
22.7
Leader
GPT-6 Astra
Updated
Nov 11, 2026
Resolution
—
Saturation
—
No signal or difficulty estimate. Held out of the fit by design.
Organizations
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | GPT-6 Astra | not reported · pi | 22.7 | |
| 2 | GPT-6.1 Terra | not reported · pi | 22.3 | |
| 3 | GPT-6.1 Sol | not reported · pi | 22.0 | |
| 4 | DeepSeek V4.1 Flash | not reported · pi | 22.0 | |
| 5 | Grok 4.6 | not reported · Grok Build | 21.2 | |
| 6 | GPT-6 Luna | not reported · pi | 20.4 | |
| 7 | Mistral Large 4 | not reported · pi | 19.4 | |
| 8 | Gemini 4 Argon | not reported · pi | 19.1 | |
| 9 | GLM-5.3 | not reported · pi | 16.5 | |
| 10 | Kimi K3 | not reported · pi | 16.5 | |
| 11 | Gemini 3.8 Flash | not reported · pi | 15.8 | |
| 12 | GPT-5.6 Sol | not reported · Codex | 15.8 | |
| 13 | GPT-5.6 Luna | not reported · pi | 15.0 | |
| 14 | Nemotron 3.5 Lightning 30B A3B (precision unspecified) | not reported · pi | 7.7 |
