DrugDiscoveryBench: leaderboard
Scale AI DrugDiscoveryBench: agentic drug-discovery research tasks spanning literature work and analysis; percent of tasks solved, best reported harness per model.
Metric: Score (%). Source: scale.com. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 51.6 | #31 |
| 2 | Gemini 3.5 Flash | 50 | #60 |
| 3 | Claude Opus 4.8 | 46.8 | #37 |
| 4 | GLM-5.2 | 36.2 | #73 |
| 5 | Kimi K2.7 Code | 35.3 | #89 |
| 6 | DeepSeek V4 Pro | 31.7 | #102 |
| 7 | Qwen 3.7 Max | 29.3 | #75 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=drugdiscoverybench · How It Works · Data refreshed daily, snapshot 2026-10-20.