DrugDiscoveryBench leaderboard
As of 6 Nov 2026, the top score on DrugDiscoveryBench is 68.7% by GPT-6 Astra (OpenAI), ahead of Gemini 4 Argon at 65.4%, across 20 model configurations with a published result (Scale AI SEAL, result dated 14 Sep 2026). Questions and tasks from the drug-discovery pipeline, from target selection to trial design, graded by domain scientists. DrugDiscoveryBench is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- GPT-6 Astra holds the top DrugDiscoveryBench score at 68.7% (Scale AI SEAL, 14 Sep 2026).
- Gemini 4 Argon is second on DrugDiscoveryBench at 65.4% (Scale AI SEAL, 22 Oct 2026).
- DrugDiscoveryBench: Drug-discovery reasoning tasks graded by practitioners. Scale AI (Scale AI SEAL).
- 20 model configurations have a published DrugDiscoveryBench result on BenchLeader (BenchLeader, 6 Nov 2026).
- DrugDiscoveryBench does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Drug-discovery reasoning tasks graded by practitioners. Scale AI.
- Published by
- Scale AI SEAL
- Category
- Knowledge
- Index weight
- Reference only
- Models
- 20
- Data as of
- 6 Nov 2026
- Newest result
- 27 Oct 2026
Scale AI SEAL Leaderboards.
What the test looks like
Questions and tasks from the drug-discovery pipeline, from target selection to trial design, graded by domain scientists.
How it is scored
Rubric score, published by Scale AI.
What to keep in mind
Expert-graded and narrow; small model set.
- 1GPT-6 Astra (max)68.7%
- 2Gemini 4 Argon (high)65.4%
- 3Muse Spark 1.3 (high)62.2%
- 4DeepSeek V4.1 Flash 1017 (xhigh)60.6%
- 5GPT-5.6 Sol (max)56.5%
- 6Claude Opus 5 (max)53.7%
- 7GPT-5.5 (xhigh)51.6%
- 8Claude Sonnet 5 (max)50.0%
- 9Gemini 3.5 Flash (high)50.0%
- 10Claude Opus 4.8 (max)46.8%
- 11Claude Sonnet 5.0 (max)44.7%
- 12Gemini 3.1 Pro (high)41.9%
- 13GLM 5.2 (xhigh)36.2%
- 14Kimi K2.7 Code (xhigh)35.3%
- 15DeepSeek V4 Pro (xhigh)31.7%
| # | ||||
|---|---|---|---|---|
| 1 |  GPT-6 Astramax | 68.7% | 67.8 | 2026-09-14 |
| 2 |  Gemini 4 Argonhigh | 65.4% | 66.0 | 2026-10-22 |
| 3 |  Muse Spark 1.3high | 62.2% | – | 2026-09-14 |
| 4 |  DeepSeek V4.1 Flash 1017xhighopen ↗ | 60.6% | – | 2026-10-27 |
| 5 | GPT-5.6 Solmax | 56.5% | 65.4 | 2026-09-11 |
| 6 |  Claude Opus 5max | 53.7% | 65.4 | 2026-09-11 |
| 7 | GPT-5.5xhigh | 51.6% | 64.3 | 2026-06-30 |
| 8 | Claude Sonnet 5max | 50.0% | 53.8 | 2026-07-14 |
| 9 | Gemini 3.5 Flashhigh | 50.0% | 62.5 | 2026-06-30 |
| 10 | Claude Opus 4.8max | 46.8% | 60.2 | 2026-06-30 |
| 11 | Claude Sonnet 5.0max | 44.7% | – | 2026-07-14 |
| 12 | Gemini 3.1 Prohigh | 41.9% | 60.1 | 2026-06-30 |
| 13 |  GLM 5.2xhighopen ↗ | 36.2% | – | 2026-06-30 |
| 14 |  Kimi K2.7 Codexhighopen ↗ | 35.3% | – | 2026-06-30 |
| 15 | DeepSeek V4 Proxhighopen ↗ | 31.7% | – | 2026-06-30 |
| 16 | Claude Sonnet 4.6max | 31.3% | 54.2 | 2026-06-30 |
| 17 | GPT-5.2xhigh | 29.3% | 60.0 | 2026-06-30 |
| 18 |  Qwen3.7xhigh | 29.3% | – | 2026-06-30 |
| 19 | Claude Opus 4.6max | 27.7% | 55.8 | 2026-06-30 |
| 20 |  MiniMax M3xhighopen ↗ | 22.8% | – | 2026-06-30 |
More knowledge benchmarks · All LLM benchmarks
Related benchmarks
Top head-to-heads on DrugDiscoveryBench
Cite as: BenchLeader, “DrugDiscoveryBench leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/scale_drugdiscovery.
DrugDiscoveryBench: questions
- What does DrugDiscoveryBench measure?
- Questions and tasks from the drug-discovery pipeline, from target selection to trial design, graded by domain scientists. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on DrugDiscoveryBench?
- As of 6 Nov 2026, the highest score on DrugDiscoveryBench is 68.7% by GPT-6 Astra (result dated 14 Sep 2026), ahead of Gemini 4 Argon at 65.4%. Scores come from Scale AI SEAL and are re-read daily.
- Which AI model leads DrugDiscoveryBench?
- GPT-6 Astra leads DrugDiscoveryBench with 68.7% as of 6 Nov 2026 (result dated 14 Sep 2026), ahead of Gemini 4 Argon at 65.4%.
- How many models have DrugDiscoveryBench results?
- 20 model configurations have a DrugDiscoveryBench result on BenchLeader, all taken from Scale AI SEAL.
- Who runs DrugDiscoveryBench and how often is it updated?
- DrugDiscoveryBench is published by Scale AI SEAL. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does DrugDiscoveryBench count toward the BenchLeader Index?
- No. DrugDiscoveryBench is shown for reference but left out of the composite index.