Category ranking · 7 models · checked 2026-10-13
Math and reasoning
Ranked by AIME 2024, pass@1. Source: Model cards (vendor-reported); sampling settings differ by vendor.
AIME 2024, pass@1
7 with a published score, 0 not yet measured.
# | Model | Parameters | Score | Source | Licence |
|---|---|---|---|---|---|
| 1 | OpenReasoning Nemotron 32B NVIDIA | 32.8 B | 89.2%Sourced | Source AIME24 column, 32B row, model card. | CC-BY-4.0 |
| 2 | DeepSeek R1-0528 Qwen3 8B DeepSeek | 8.2 B | 86%Sourced | Source AIME 24 column, model card table. | MIT |
| 3 | QwQ 32B Alibaba Qwen | 32.8 B | 79.5%Sourced | Source Third-party report: the QwQ 32B row in Microsoft's Phi-4-reasoning model card. | Apache-2.0 |
| 4 | Phi-4 reasoning Microsoft | 14.7 B | 75.3%Sourced | Source AIME 24 column, model card table. | MIT |
| 5 | OpenMath Nemotron 7B NVIDIA | 7.6 B | 74.8%Sourced | Source AIME24, chain-of-thought pass@1 (the card also gives maj@64 in brackets). | CC-BY-4.0 |
| 6 | Phi-4 mini reasoning Microsoft | 3.8 B | 57.5%Sourced | Source "AIME" column of the model card (AIME 2024 per the linked paper). | MIT |
| 7 | DeepSeek R1 Distill Qwen 7B DeepSeek | 7.6 B | 55.5%Sourced | Source Distilled-model evaluation table, model card. | MIT |
How to read this: one public benchmark per category, so you can compare like with like. We copy the published number and link it; we do not re-run it, average different benchmarks or fill gaps. Vendor-reported numbers use each vendor's own settings, so treat small gaps as ties. “Not yet measured” means we found no comparable public score, not that the model is weak.
Every model was checked against the Hugging Face API on 2026-10-13: it exists, is public, and has the parameter count and licence shown. Other categories: Coding · Vision-language · OCR and documents · Speech to text · Text to speech · Image generation · Embeddings and rerankers · Small fast 'System 1' · Biomedical · Tool use and agents · Multilingual and translation · Time series · Robotics (VLA).
Machine-readable: /api/v1/categories/math-reasoning.json