MolViBench - Level 5 Create: leaderboard
Metric: Pass@1 (%) on the 64 Level 5 tasks (Create), 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 | GPT-5.2 Codex | 9.4 |
| 2 | Gemini 3 Pro | 7.8 |
| 3 | DeepSeek V3.2 (Non-reasoning) | 7.8 |
| 4 | Claude Opus 4.6 | 6.2 |
| 5 | DeepSeek V3.2 (Thinking) | 6.2 |
| 6 | Claude Opus 4.6 (Thinking) | 6.2 |
| 7 | GPT-5.3 Codex | 6.2 |
| 8 | Kimi K2.5 | 4.7 |
| 9 | MiniMax-M2.5 | 3.1 |
Interactive version: theaggregate.ai/benchmark?slug=molvibench-level-5-create · How It Works · Data refreshed daily, snapshot 2026-10-07.