MolecularIQ - Generation: leaderboard
Measures symbolic reasoning over molecular graphs in the MolecularIQ paper snapshot: feature counting, atom/bond indexing and constrained molecule generation. The paper reports average accuracy over three rollouts, not pass@3. Chemistry fine-tunes, dated checkpoints and reasoning effort remain distinct. One source Qwen-2.5 14B checkpoint is unspecified and is explicitly labelled rather than assigned to a known base/instruct model. This series reports MolecularIQ - Generation using Mean Accuracy (%).
Metric: Mean Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B (High) | 53.7 | #346 (GPT-OSS-120B) |
| 2 | Qwen 3 235B A22B 2507 FP8 (Thinking) | 46.7 | #253 |
| 3 | Qwen 3 Next 80B A3B (Thinking) | 40 | #312 |
| 4 | GPT-OSS-20B (High) | 35.2 | #523 (GPT-OSS-20B) |
| 5 | GPT-OSS-20B (Medium) | 32.1 | #523 (GPT-OSS-20B) |
| 6 | Qwen 3 30B A3B 2507 (Thinking) | 29.4 | #379 |
| 7 | GPT-OSS-120B (Medium) | 28 | #346 (GPT-OSS-120B) |
| 8 | GLM-4.6 FP8 | 22 | #269 |
| 9 | Qwen 3 32B | 21 | #449 |
| 10 | GPT-OSS-120B (Low) | 20.3 | #346 (GPT-OSS-120B) |
| 11 | Seed-OSS-36B-Instruct | 20 | #384 |
| 12 | Qwen 3 14B | 17.7 | #542 |
| 13 | Qwen 3 8B | 15.8 | #678 |
| 14 | NVIDIA Nemotron Nano 9B V2 | 14.8 | #855 |
| 15 | Mistral Small 3 | 14.1 | #630 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=moleculariq-generation · How It Works · Data refreshed daily, snapshot 2026-11-05.