MolViBench - Level 2 Apply: leaderboard

The Level 2 slice of MolViBench (2026), a molecular vibe coding benchmark: RDKit tasks on applying: multi-tool composition with two-step molecular transformations (72 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 72 Level 2 tasks (Apply), 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 2027. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.3 Codex55.6#125
2Claude Opus 4.6Thinking50#114
3Gemini 3 Pro45.8#146
4GPT-5.2 Codex44.4#186
5Kimi K2.543.1#272
6MiniMax-M2.540.3#600
7DeepSeek V3.2Non-reasoning40.3#425

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-2-apply · How It Works · Data refreshed daily, snapshot 2026-11-22.