# American Invitational Mathematics Examination 2025 (AIME 2025)

> The most recent AIME examination, featuring 15 challenging mathematics problems testing olympiad-level mathematical reasoning with integer answers from 000-999.

Canonical page: https://benchlm.ai/benchmarks/aime2025

- Category: [Mathematics](/math)
- Last updated: October 17, 2026

## About AIME 2025

- Year: 2025
- Tasks: 15 problems
- Format: Integer answers 000-999
- Difficulty: High school olympiad level
- Paper: [American Invitational Mathematics Examination](https://www.maa.org/math-competitions/aime)

AIME 2025 represents the current standard for intermediate-level mathematical olympiad problems. Success requires sophisticated mathematical reasoning and problem-solving techniques.

AIME 2025 is currently displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Leaderboard (51 models)

| Rank | Model | Creator | Score |
|------|-------|---------|-------|
| 1 | [GPT-5.2](/models/gpt-5-2) | OpenAI | 100% |
| 2 | [Step 3.5 Flash](/models/step-3-5-flash) | StepFun | 97.3% |
| 3 | [MAI-Thinking-1](/models/mai-thinking-1) | Microsoft | 97% |
| 4 | [Kimi K2.5 (Reasoning)](/models/kimi-k2-5-reasoning) | Moonshot AI | 96.1% |
| 5 | [Kimi K2.5](/models/kimi-k2-5) | Moonshot AI | 96.1% |
| 6 | [GLM-4.7](/models/glm-4-7) | Z.AI | 95.7% |
| 7 | [Ternary Bonsai 2 27B](/models/ternary-bonsai-2-27b) | Prism ML | 95% |
| 8 | [MiMo-V2-Flash](/models/mimo-v2-flash) | Xiaomi | 94.1% |
| 9 | [DeepSeek V3.2 (Thinking)](/models/deepseek-v3-2-thinking) | DeepSeek | 93.1% |
| 10 | [o4-mini (high)](/models/o4-mini-high) | OpenAI | 92.7% |
| 11 | [GPT-OSS 120B](/models/gpt-oss-120b) | OpenAI | 92.5% |
| 12 | [Qwen3 235B 2507 (Reasoning)](/models/qwen3-235b-2507-reasoning) | Alibaba | 92.3% |
| 13 | [GPT-OSS 20B](/models/gpt-oss-20b) | OpenAI | 91.7% |
| 14 | [GLM-4.7-Flash](/models/glm-4-7-flash) | Z.AI | 91.6% |
| 15 | [Nemotron 3 Super 100B](/models/nemotron-3-super-100b) | NVIDIA | 90.2% |
| 16 | [Granite 4.2 30B](/models/granite-4-2-30b) | IBM | 89.2% |
| 17 | [Nemotron 3 Nano 30B](/models/nemotron-3-nano-30b) | NVIDIA | 89.1% |
| 18 | [DeepSeek V3.1 (Reasoning)](/models/deepseek-v3-1-reasoning) | DeepSeek | 88.4% |
| 19 | [Sarvam 105B](/models/sarvam-105b) | Sarvam | 88.3% |
| 20 | [DeepSeek-R1-0528](/models/deepseek-r1-0528) | DeepSeek | 87.5% |
| 21 | [Claude Sonnet 4.5](/models/claude-sonnet-4-5) | Anthropic | 87% |
| 22 | [Granite 4.2 8B](/models/granite-4-2-8b) | IBM | 86.7% |
| 23 | [MiniCPM5-2B](/models/minicpm5-2b) | OpenBMB | 86.5% |
| 24 | [Exaone 4.0 32B](/models/exaone-4-0-32b) | LG AI Research | 85.3% |
| 25 | [Ministral 3 14B (Reasoning)](/models/ministral-3-14b-reasoning) | Mistral | 85% |
| 26 | [Nemotron 3 Nano Omni 30B A3B](/models/nemotron-3-nano-omni-30b-a3b) | NVIDIA | 82.1% |
| 27 | [Qwen3 235B-A22B (Reasoning)](/models/qwen3-235b-a22b-reasoning) | Alibaba | 81.5% |
| 28 | [Granite 4.2 3B](/models/granite-4-2-3b) | IBM | 78.3% |
| 29 | [Phi-4-reasoning-plus](/models/phi-4-reasoning-plus) | Microsoft | 78% |
| 30 | [MiniMax M1 80k](/models/minimax-m1-80k) | MiniMax | 76.9% |
| 31 | [Claude 4 Opus Thinking](/models/claude-4-opus-thinking) | Anthropic | 75.5% |
| 32 | [Qwen3 32B (Reasoning)](/models/qwen3-32b-reasoning) | Alibaba | 72.9% |
| 33 | [Nemotron Ultra 253B](/models/nemotron-ultra-253b) | NVIDIA | 72.5% |
| 34 | [Qwen3 235B 2507](/models/qwen3-235b-2507) | Alibaba | 70.3% |
| 35 | [QwQ-32B](/models/qwq-32b) | Alibaba | 69.5% |
| 36 | [Magistral Medium](/models/magistral-medium) | Mistral | 64.9% |
| 37 | [Phi-4-reasoning](/models/phi-4-reasoning) | Microsoft | 62.9% |
| 38 | [Magistral Small](/models/magistral-small) | Mistral | 62.8% |
| 39 | [LFM2.5-2.6B](/models/lfm2-5-2-6b) | LiquidAI | 51.9% |
| 40 | [DeepSeek V3.1](/models/deepseek-v3-1) | DeepSeek | 49.8% |
| 41 | [Gemini 2.5 Flash-Lite](/models/gemini-2-5-flash-lite) | Google | 49.8% |
| 42 | [Kimi K2](/models/kimi-k2) | Moonshot AI | 49.5% |
| 43 | [LFM2.5-8B-A1B](/models/lfm2-5-8b-a1b) | LiquidAI | 42.5% |
| 44 | [MiniCPM5-1B](/models/minicpm5-1b) | OpenBMB | 40.4% |
| 45 | [Claude 4 Opus](/models/claude-4-opus) | Anthropic | 33.9% |
| 46 | [Claude 4 Sonnet](/models/claude-4-sonnet) | Anthropic | 33.1% |
| 47 | [Gemini 2.0 Flash](/models/gemini-2-0-flash) | Google | 29.7% |
| 48 | [Qwen3 235B-A22B](/models/qwen3-235b-a22b) | Alibaba | 24.7% |
| 49 | [Gemini 2.0 Flash-Lite](/models/gemini-2-0-flash-lite) | Google | 23.8% |
| 50 | [Qwen3 32B](/models/qwen3-32b) | Alibaba | 20.2% |
| 51 | [Gemini 1.5 Flash](/models/gemini-1-5-flash) | Google | 14.7% |

## FAQ

### What does AIME 2025 measure?

The most recent AIME examination, featuring 15 challenging mathematics problems testing olympiad-level mathematical reasoning with integer answers from 000-999.

### Which model scores highest on AIME 2025?

GPT-5.2 by OpenAI currently leads with a score of 100% on AIME 2025.

### How many models are evaluated on AIME 2025?

51 AI models have published results on AIME 2025 in the BenchLM catalog.

### Does AIME 2025 affect BenchLM's overall score?

Not directly. AIME 2025 is still displayed on BenchLM for reference, but it is excluded from the weighted scoring formula.

## Compare Top Models on AIME 2025

- [GPT-5.2 vs Step 3.5 Flash](/compare/gpt-5-2-vs-step-3-5-flash)
- [Step 3.5 Flash vs MAI-Thinking-1](/compare/mai-thinking-1-vs-step-3-5-flash)
- [MAI-Thinking-1 vs Kimi K2.5 (Reasoning)](/compare/kimi-k2-5-reasoning-vs-mai-thinking-1)
- [Kimi K2.5 (Reasoning) vs Kimi K2.5](/compare/kimi-k2-5-vs-kimi-k2-5-reasoning)

## Related Reading

- [AIME 2025 benchmark explainer](/blog/posts/aime-hmmt-competition-math)