MolViBench - Level 4 Analyze and Evaluate: leaderboard
The Level 4 slice of MolViBench (2026), a molecular vibe coding benchmark: RDKit tasks on analyzing and evaluating: branching, iteration and multi-step decisions over molecules (75 tasks). The score is Pass@1, the percent of tasks whose generated Python code reproduces the reference output, with each model generating once without execution feedback; higher is better.
Metric: Pass@1 (%) on the 75 Level 4 tasks (Analyze and Evaluate), 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 October 2028. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Opus 4.6 | 24 | #114 | |
| 2 | Gemini 3 Pro | 21.3 | #146 | |
| 3 | MiniMax-M2.5 | 14.7 | #600 | |
| 4 | DeepSeek V3.2 | Thinking | 13.3 | #425 |
| 5 | GPT-5.2 Codex | 9.3 | #186 | |
| 6 | Kimi K2.5 | 6.7 | #272 | |
| 7 | GPT-5.3 Codex | 5.3 | #125 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=molvibench-level-4-analyze-and-evaluate · How It Works · Data refreshed daily, snapshot 2026-11-22.