AIME 2025 leaderboard
As of November 14, 2026, among the results listed here, MoonshotAI Kimi K2 Thinking has the highest reported AIME 2025 score, 100.0%, published by Moonshot AI. 11 models have a AIME 2025 result in our records.
Scores as the model makers published them, not re-run by us; each row links to its announcement.
11 models ranked · 5 makers
| Rank | Model | Score | Setting | Reported by | Date |
|---|---|---|---|---|---|
| 1 | Kimi K2 Thinking | 100.0% | Not published | Moonshot AI | November 6, 2025 (model release) |
| 1 | GPT-5.2 Pro | 100.0% | no tools | OpenAI | December 10, 2025 (model release) |
| 3 | Kimi K2.5 | 96.1% | Thinking | Moonshot AI | January 27, 2026 (model release) |
| 4 | GLM 4.7 | 95.7% | Not published | Z.ai | December 22, 2025 (model release) |
| 5 | GLM 4.6V | 93.9% | Not published | Z.ai | December 22, 2025 (model release) |
| 6 | gpt-oss-120b | 92.5% | no tools | OpenAI | August 5, 2025 (model release) |
| 7 | gpt-oss-20b | 91.7% | no tools | OpenAI | August 5, 2025 (model release) |
| 8 | Mistral Small 4 | 84.0% | Reasoning | Mistral AI | March 16, 2026 (model release) · read from the maker's chart |
| 9 | Kimi K2 0905 | 75.2% | Not published | Moonshot AI | November 6, 2025 (model release) |
| 10 | Kimi K2 0711 | 49.5% | Instruction model evaluation results; Avg@64 | Moonshot AI | July 11, 2025 (model release) |
| 11 | H SmolLM3-3B | 36.7% | Instruction Model; Extended Thinking | HuggingFaceTB | Publication date not reported |
Other reported runs
How much to trust these numbers
- Source
- Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
- Settings
- Settings differ between makers. Check the Setting column before comparing two scores.
- Spread
- 100.0% to 36.7% across 11 models.
- Freshness
- Newest result published March 16, 2026; oldest July 11, 2025.
- Charts
- Some values were read from charts in the announcements. Two independent readings had to match exactly.
What AIME 2025 measures
Tests how well models solve all 30 problems from the 2025 American Invitational Mathematics Examination; the results do not specify how scores are calculated.