FigQA: leaderboard
181 multiple-choice questions in the FigQA subtask of FutureHouse's LAB-Bench (2024), each pairing a figure from a biology research paper with a question answerable only by reading that figure.
Metric: Score (system card). Source: arxiv.org. Source: benchmarklist.com. Saturation forecast: Around October 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Mythos 5 | 90.7 | #2 |
| 2 | Claude Mythos Preview | 89.3 | #4 |
| 3 | Claude Opus 4.8 | 87.3 | #37 |
| 4 | Claude Opus 4.7 | 85.4 | #49 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=figqa · How It Works · Data refreshed daily, snapshot 2026-10-20.