ReactBench (Reaction Diagrams) - Reasoning: leaderboard
Metric: Answer accuracy (%) on the structural topology reasoning questions (classify the global topology, detect cycles and terminal nodes), over the 1,618 expert-annotated question-answer pairs on more than 1,300 real chemical reaction diagrams from chemistry journals and patents of ReactBench (single-line, multi-line, tree and graph topologies), multiple-choice or short open answers matched exactly against the key after template extraction, no LLM judge; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 64.85 |
| 2 | Qwen 3 VL 32B Instruct | 57.43 |
| 3 | Gemini 1.5 Pro | 55.45 |
| 4 | GPT-5.5 | 51.49 |
| 5 | GPT-4o | 49.5 |
| 6 | Claude 3.5 Sonnet | 49.5 |
| 7 | Claude Sonnet 4.6 | 46.53 |
| 8 | Qwen 3 VL 32B (Thinking) | 43.07 |
| 9 | Qwen 2.5 VL 7B | 39.11 |
| 10 | Qwen 3 VL 8B Instruct | 32.67 |
| 11 | Qwen 3 VL 8B (Thinking) | 29.11 |
| 12 | InternVL2.5-78B | 26.73 |
Interactive version: theaggregate.ai/benchmark?slug=reactbench-reaction-diagrams-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.