Atlas
Benchmarks
326 sources · 1,256 models × 2,956 benchmarks · 154,484 scoresScience's Last Exam: Hypothesis Discovery
Whether a model can formulate valuable, testable hypotheses from real data, over 5 tasks in the DiscoveryRSI research harness. The inspire rate is the share of its hypothesis versions that anticipate research published after its knowledge cutoff.
Models
6
Top score
12.4
Leader
MiMo-V2.6-Pro
Updated
Nov 4, 2026
Resolution
—
Saturation
—
No signal or difficulty estimate. Too few models have been measured on this benchmark to estimate its difficulty.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | MiMo-V2.6-Pro | not reported · Science's Last Exam; Claude Code with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 12.4 | |
| 2 | GLM-5.3 Flash | high · Science's Last Exam; Claude Code with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 11.6 | |
| 3 | Kimi K3.1 Pro | not reported · Science's Last Exam; Claude Code with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 9.2 | |
| 4 | Claude Opus 5 | high · Science's Last Exam; Claude Code with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 7.3 | |
| 5 | GPT-6.1 Sol | medium · Science's Last Exam; Codex with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 4.9 | |
| 6 | MiniMax M3 | high · Science's Last Exam; Claude Code with web search and fetch off; DiscoveryRSI for Hypothesis Discovery | 3.6 |
