Quantitative reasoning and autonomous analysis on BixBench Human Verified-50
83.33AccuracyCellType
Evaluation results
| Method | Links | |
|---|---|---|
| CellTypeCore model=Claude Opus 4.72026.05 | 83.33 | |
| Claude CodeCore model=Claude Opus 4.72026.05 | 83.33 | |
| BIORESEARCHERCore model=GPT-5.4, Reasoning effort=medium2026.05 | 81.82 |