ChemPro (MCQ) - Difficult: leaderboard
Metric: Multiple-choice accuracy (%) on the Difficult section (JEE Mains competitive-exam questions, 2020-2024) of ChemPro; pass@1, mean of five runs, fixed system and wraparound prompts for every model; printed as fractions and shown times 100; Table 2 lists only the three best open models of each size class (7B, 10B, 14B, 32B, 70B) and OpenAI o1-mini, o3-mini and o1; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 | 76 | #177 |
| 2 | O3 Mini | 75 | #266 |
| 3 | O1 Mini | 75 | #313 |
| 4 | Rombos-LLM-V2.5-Qwen-32B | 69 | #353 |
| 5 | Qwen2.5-32B-Instruct-abliterated-v2 | 68 | #381 |
| 6 | Lamarckvergence-14B | 63 | #533 |
| 7 | Falcon3-10B-Instruct | 62 | #1142 |
| 8 | Falcon3-7B-Instruct | 58 | #1284 |
| 9 | HomerCreativeAnvita-Mix-Qw7B | 53 | #917 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=chempro-mcq-difficult · How It Works · Data refreshed daily, snapshot 2026-10-11.