MathNet-RAG - Zero-Shot: leaderboard
MathNet-RAG (2026) gives 35 Olympiad problems, each paired by experts with a related problem, and human experts grade the solutions. Accuracy when the model sees only the target problem, in percent; higher is better.
Metric: on MathNet-RAG, 35 Olympiad problems each paired by experts with a related problem, graded by human experts; accuracy (%) with only the target problem; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 89.1 | #123 |
| 2 | DeepSeek V3.2 Speciale | 84.8 | #277 |
| 3 | GPT-5 | 76.8 | #185 |
| 4 | Grok 4.1 Fast | 75.4 | #431 |
| 5 | Claude Opus 4.5 | 46.8 | #165 |
| 6 | OLMo 3 32B (Thinking) | 45.2 | |
| 7 | Phi-4-reasoning-plus | 15.1 | #2288 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=mathnet-rag-zero-shot · How It Works · Data refreshed daily, snapshot 2026-11-22.