American Invitational Mathematics Examination 2025 (AIME 2025)
We show this table for reference; we do not rank on it.
The most recent AIME examination, featuring 15 challenging mathematics problems testing olympiad-level mathematical reasoning with integer answers from 000-999.
Benchmark score on AIME 2025 — October 17, 2026
We compile the AIME 2025 rows from provider self-reports and secondary reports. GPT-5.2 leads the table at 100%, followed by Step 3.5 Flash (97.3%) and MAI-Thinking-1 (97%). We do not use these results to rank models overall.
GPT-5.2
OpenAI
Step 3.5 Flash
StepFun
MAI-Thinking-1
Microsoft
53 modelsMathematicsCurrentDisplay onlyUpdated October 17, 2026
Benchmark score table (53 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-5.2OpenAI | 100% | Not reported | Closed |
| 2 | Step 3.5 FlashStepFun | 97.3% | Not reported | Open |
| 3 | MAI-Thinking-1Microsoft | 97% | Not reported | Closed |
| 4 | Kimi K2.5 (Reasoning)Moonshot AI | 96.1% | Not reported | Closed |
| 5 | Kimi K2.5Moonshot AI | 96.1% | Not reported | Open |
| 6 | GLM-4.7Z.AI | 95.7% | Not reported | Open |
| 7 | Ternary Bonsai 2 27BPrism ML | 95% | Not reported | Open |
| 8 | MiMo-V2-FlashXiaomi | 94.1% | Not reported | Open |
| 9 | DeepSeek V3.2 (Thinking)DeepSeek | 93.1% | Not reported | Open |
| 10 | o4-mini (high)OpenAI | 92.7% | Not reported | Closed |
| 11 | GPT-OSS 120BOpenAI | 92.5% | Not reported | Open |
| 12 | Qwen3 235B 2507 (Reasoning)Alibaba | 92.3% | Not reported | Open |
| 13 | GPT-OSS 20BOpenAI | 91.7% | Not reported | Open |
| 14 | GLM-4.7-FlashZ.AI | 91.6% | Not reported | Open |
| 15 | Mercury 2Inception | 91.1% | Not reported | Closed |
| 16 | Nemotron 3 Super 100BNVIDIA | 90.2% | Not reported | Open |
| 17 | Granite 4.2 30BIBM | 89.2% | Not reported | Open |
| 18 | Nemotron 3 Nano 30BNVIDIA | 89.1% | Not reported | Open |
| 19 | DeepSeek V3.1 (Reasoning)DeepSeek | 88.4% | Not reported | Open |
| 20 | Sarvam 105BSarvam | 88.3% | Not reported | Open |
| 21 | DeepSeek-R1-0528DeepSeek | 87.5% | Not reported | Open |
| 22 | Claude Sonnet 4.5Anthropic | 87% | Not reported | Closed |
| 23 | Granite 4.2 8BIBM | 86.7% | Not reported | Open |
| 24 | MiniCPM5-2BOpenBMB | 86.5% | Not reported | Open |
| 25 | Exaone 4.0 32BLG AI Research | 85.3% | Not reported | Open |
| 26 | GPT-5 nanoOpenAI | 85.2% | Not reported | Closed |
| 27 | Ministral 3 14B (Reasoning)Mistral | 85% | Not reported | Open |
| 28 | Nemotron 3 Nano Omni 30B A3BNVIDIA | 82.1% | Not reported | Open |
| 29 | Qwen3 235B-A22B (Reasoning)Alibaba | 81.5% | Not reported | Open |
| 30 | Granite 4.2 3BIBM | 78.3% | Not reported | Open |
| 31 | Phi-4-reasoning-plusMicrosoft | 78% | Not reported | Open |
| 32 | MiniMax M1 80kMiniMax | 76.9% | Not reported | Closed |
| 33 | Claude 4 Opus ThinkingAnthropic | 75.5% | Not reported | Closed |
| 34 | Qwen3 32B (Reasoning)Alibaba | 72.9% | Not reported | Open |
| 35 | Nemotron Ultra 253BNVIDIA | 72.5% | Not reported | Open |
| 36 | Qwen3 235B 2507Alibaba | 70.3% | Not reported | Open |
| 37 | QwQ-32BAlibaba | 69.5% | Not reported | Open |
| 38 | Magistral MediumMistral | 64.9% | Not reported | Closed |
| 39 | Phi-4-reasoningMicrosoft | 62.9% | Not reported | Open |
| 40 | Magistral SmallMistral | 62.8% | Not reported | Open |
| 41 | LFM2.5-2.6BLiquidAI | 51.9% | Not reported | Open |
| 42 | DeepSeek V3.1DeepSeek | 49.8% | Not reported | Open |
| 43 | Gemini 2.5 Flash-LiteGoogle | 49.8% | Not reported | Closed |
| 44 | Kimi K2Moonshot AI | 49.5% | Not reported | Closed |
| 45 | LFM2.5-8B-A1BLiquidAI | 42.5% | Not reported | Open |
| 46 | MiniCPM5-1BOpenBMB | 40.4% | Not reported | Open |
| 47 | Claude 4 OpusAnthropic | 33.9% | Not reported | Closed |
| 48 | Claude 4 SonnetAnthropic | 33.1% | Not reported | Closed |
| 49 | Gemini 2.0 FlashGoogle | 29.7% | Not reported | Closed |
| 50 | Qwen3 235B-A22BAlibaba | 24.7% | Not reported | Open |
| 51 | Gemini 2.0 Flash-LiteGoogle | 23.8% | Not reported | Closed |
| 52 | Qwen3 32BAlibaba | 20.2% | Not reported | Open |
| 53 | Gemini 1.5 FlashGoogle | 14.7% | Not reported | Closed |
Among the reported AIME 2025 rows, GPT-5.2 is first at 100%. The third row is 3.0 points behind. The broader top-10 range is 7.3 points, so many of the published results sit in a relatively narrow band.
53 models have been evaluated on AIME 2025. The benchmark falls in the Mathematics category. AIME 2025 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About AIME 2025
Year
2025
Tasks
15 problems
Format
Integer answers 000-999
Difficulty
High school olympiad level
AIME 2025 represents the current standard for intermediate-level mathematical olympiad problems. Success requires sophisticated mathematical reasoning and problem-solving techniques.
Freshness and provenance
Version
AIME 2025
Refresh cadence
Annual
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
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?
53 AI models have published results on AIME 2025 in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 5,600+ readers.
One email each week. Unsubscribe anytime.