AI benchmarks and leaderboards are different things
Learn the difference between a benchmark, an evaluation run and a leaderboard, and use each to answer the right question.
Model reports
AIME
pass@k, majority vote and Python tool access each shift this by double digits.
Evaluation context and evidence
- Source
- Model reports
- Publisher
- Not recorded
- Benchmark release
- 2024-02-01
- Collected
- Not recorded
- Modality
- Not recorded
These are source-reported observations, not independent reruns by this site. Check the task version, evaluation protocol, unit and reporting date before comparing models. Results from different setups are not a universal ranking.
Original sources
Reported results 10
Shown in source observation order, not ranked. Dates retain the precision provided by the source.
| Model | Value | Unit | Protocol | Reported date and basis | Source |
|---|---|---|---|---|---|
| Claude Opus 4 | 75.5 | percent | extended thinking up to 64K tokens, top-p 0.95, no parallel test-time compute | 2025-05-22 · document_publication | Claude Opus 4 and Sonnet 4 |
| Claude Sonnet 4 | 70.5 | percent | extended thinking up to 64K tokens, top-p 0.95, no parallel test-time compute | 2025-05-22 · document_publication | Claude Opus 4 and Sonnet 4 |
| DeepSeek-R1 | 79.8 | percent | Pass@1 | 2025-01-22 · document_publication | DeepSeek-R1 |
| DeepSeek-V3 | 39.2 | percent | Pass@1 | 2024-12-27 · document_publication | DeepSeek-V3 |
| Gemma 4 (31B) | 89.2 | percent | no tools | 2026-03-11 · document_publication | Gemma 4 (31B) |
| Qwen3.5-397B-A17B | 91.3 | percent | thinking (temp 0.6, top_p 0.95, top_k 20) | 2026-02-16 · document_publication | Qwen3.5-397B-A17B |
| Qwen3-235B-A22B (Thinking) | 85.7 | percent | thinking mode | 2025-05-14 · document_publication | Qwen3 |
| GLM-5.1 | 95.3 | percent | AIME 2026, as reported | 2026-04-03 · document_publication | GLM-5.1 |
| GLM-5.2 | 99.2 | percent | AIME 2026, temp 1.0, top_p 0.95, max gen 163840 tokens, GPT-5.5 (medium) judge | 2026-06-16 · document_publication | GLM-5.2 |
| GLM-5 | 92.7 | percent | AIME 2026 I, temp 1.0, top_p 0.95, GPT-5.2 (medium) judge | 2026-02-11 · document_publication | GLM-5 |