OpenAI MedChemBench: leaderboard
OpenAI's MedChemBench (20 November 2026): 99 problems built from 117 medicinal chemistry papers and patents, where a model gets a drug program's target, goals, compounds and results so far and must reconstruct structures, explain findings, pick compounds and predict properties. One attempt per response, no tools, ten responses per problem. This is the overall score across all four graded components; it also carries GPT-6.2 Astra, which the component charts leave out. Rows are model and reasoning-effort pairs for twelve OpenAI models; GPT-5.2, GPT-5.4 and GPT-5.5 come from the paper's tables. Share of rubric credit (%); higher is better.
Metric: Average credit (%). Source: openai.com. Saturation forecast: Around February 2027. 12 models tracked. Updated 21 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-6.2 Astra | Max | 85.6 | #4 |
| 2 | GPT-6 Astra | Max | 83.83 | #10 |
| 3 | GPT-6.1 Sol | Max | 82.89 | #15 |
| 4 | GPT-6.1 Terra | Max | 77.56 | #25 |
| 5 | GPT-6 Sol | Max | 69.07 | #34 |
| 6 | GPT-5.6 Sol | Max | 59.77 | #37 |
| 7 | GPT-5.5 | xHigh | 49.26 | #54 |
| 8 | GPT-5.6 Terra | Max | 43.26 | #71 |
| 9 | GPT-5.6 Luna | Max | 40.57 | #131 |
| 10 | GPT-6 Luna | Max | 40.1 | #105 |
| 11 | GPT-5.4 | xHigh | 39.24 | #135 |
| 12 | GPT-5.2 | xHigh | 37.83 | #185 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #7 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=openai-medchembench · How It Works · Data refreshed daily, snapshot 2026-11-21.