SupraBench - Solvent Identification (CoT): leaderboard
Solvent identification in SupraBench (2026), a supramolecular chemistry benchmark: macro-F1 times 100 for naming which of six solvents a binding constant was measured in, with a chain-of-thought prompt. Higher is better.
Metric: Macro-F1 x100 (0-100) of six-way identification of the solvent in which a binding constant was measured (2,172 records), chain-of-thought prompt (explicit reasoning before the answer). Source: arxiv.org. Saturation forecast: Around March 2028. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 33.1 | #164 |
| 2 | Qwen 3.5 27B | 25.4 | #359 |
| 3 | GPT-5.4 Mini | 22 | #389 |
| 4 | GPT-5.4 Nano | 17.4 | #635 |
| 5 | Qwen 3.5 9B | 16.7 | #803 |
| 6 | Llama 3.1 8B Instruct | 15.4 | #4246 |
| 7 | Llama 3.1 70B Instruct | 10.6 | #1546 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=suprabench-solvent-identification-cot · How It Works · Data refreshed daily, snapshot 2026-11-13.