Benchmark
MedChemBench
Science benchmarks measure how well models answer science questions. Scores are percentages. Higher is better.
- Models measured
- 5
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 41%
- Mean
- 47.4%
- Median
- 7.5%
- Std dev
- 35%
- 10th percentile
- 49.3%
- 90th percentile
- 30.4%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | GPT-6 Astrainternal · weighted score· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 49.3% |
| 2 | GPT-5.6 Solinternal· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 48.3% |
| 3 | GPT-5.6 Solinternal · weighted score· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 47.4% |
| 4 | GPT 5.5internal· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 35.5% |
| 5 | GPT-5.6 Terrainternal· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 35% |
| 6 | GPT-5.6 Lunainternal· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 30.4% |