MolViBench - Level 3 Analyze: leaderboard
Metric: Pass@1 (%) on the 72 Level 3 tasks (Analyze), direct generation in one pass with no execution feedback, RDKit molecular coding tasks answered by generated Python code, official APIs, temperature 0, at most 4,096 output tokens; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.6 (Thinking) | 30.6 |
| 2 | Claude Opus 4.6 | 27.8 |
| 3 | MiniMax-M2.5 | 26.4 |
| 4 | GPT-5.2 Codex | 23.6 |
| 5 | Gemini 3 Pro | 19.4 |
| 6 | DeepSeek V3.2 (Thinking) | 19.4 |
| 7 | GPT-5.3 Codex | 16.7 |
| 8 | DeepSeek V3.2 (Non-reasoning) | 15.3 |
| 9 | Kimi K2.5 | 6.9 |
Interactive version: theaggregate.ai/benchmark?slug=molvibench-level-3-analyze · How It Works · Data refreshed daily, snapshot 2026-10-07.