MolDesignBench - Molecule Generation (Without Tools): leaderboard
Metric: Success rate (%; share of the 1,000 generation instances handled correctly: a valid SMILES satisfying every implicit and explicit property and functional-group constraint on the 900 feasible instances, or an explicit None on the 100 infeasible ones; agent without tools, from the model's own reasoning, at most 20 turns). Source: arxiv.org. Saturation forecast: Around 2034. 6 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 11 |
| 2 | Qwen 3 235B A22B (Thinking) | 8 |
| 3 | Intern-S1 | 8 |
| 4 | GPT-OSS-120B | 6 |
| 5 | Claude Opus 4.6 (Thinking) | 4 |
Interactive version: theaggregate.ai/benchmark?slug=moldesignbench-molecule-generation-without-tools · How It Works · Data refreshed daily, snapshot 2026-09-26.