SMDD-Bench Diversity: leaderboard
SMDD-Bench diversity slice measuring whether agents generate multiple distinct, novel, successful molecule designs across repeated Lead Optimization rollouts.
Metric: Avg Successful. Source: smddbench.com. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 8.4 | #90 |
| 2 | Gemini 3.1 Pro (Preview) | 8 | #58 |
| 3 | GPT-5.4 (Medium) | 7.9 | #79 (GPT-5.4) |
| 4 | Qwen 3.5 397B A17B | 7.25 | #146 |
| 5 | MiniMax-M2.7 | 6 | #258 |
| 6 | Kimi K2.5 (Thinking) | 6 | #144 (Kimi K2.5) |
| 7 | DeepSeek V3.2 | 5.35 | #205 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=smdd-bench-diversity · How It Works · Data refreshed daily, snapshot 2026-10-20.