AIME 2025
Olympiad mathematics from the 2025 American Invitational Mathematics Examination.
MathematicsMathematical Association of America208 models run by Artificial AnalysisUpdated 30 September 2026
How it is run
Accuracy on 30 problems with integer answers, using AA’s published model configurations.
What it cannot tell you
Legacy, small and near-saturated for leading models; published problems can enter training data.
- Leader
- GPT-5.299.00% · Artificial Analysis results, grouped here by release
- Top-ten spread
- 4.67%Leader minus tenth place, or last place in smaller fields.
- Median
- 52.33%of the 208 rows published
- No published result
- 46of 46 models on this site
Artificial Analysis results by release
Highest published effort per model release, grouped here from AA’s public results. Grouping ranks are derived here; source settings and model-publisher reports stay separate.
| # | Model | Score | Effort |
|---|---|---|---|
| 1 | GPT-5.2 | 99.00% | xhigh |
| 2 | GPT-5 Codex | 98.67% | high |
| 3 | DeepSeek V3.2 Speciale | 96.67% | Not specified |
| 4 | GPT-5.1 Codex | 95.67% | high |
| 4 | Gemini 3 Pro Preview | 95.67% | high |
| 6 | GLM-4.7 | 95.00% | Not specified |
| 7 | Kimi K2 Thinking | 94.67% | Not specified |
| 7 | KAT-Coder-Pro V1 | 94.67% | Not specified |
| 9 | GPT-5 | 94.33% | high |
| 9 | Nova 2.0 Lite | 94.33% | high |
| 11 | GPT-5.1 | 94.00% | high |
| 12 | gpt-oss-120b | 93.44% | high |
| 13 | Grok 4 | 92.67% | Not specified |
| 14 | GPT-5.1 Codex mini | 91.67% | high |
| 15 | Claude Opus 4.5 | 91.33% | Not specified |
| 16 | Qwen3 235B A22B 2507 (Reasoning) | 91.00% | Not specified |
| 17 | GPT-5 mini | 90.67% | high |
| 17 | o4-mini | 90.67% | high |
| 19 | K-EXAONE | 90.33% | Not specified |
| 20 | Nova 2.0 Omni | 89.67% | medium |
| 20 | DeepSeek V3.1 Terminus | 89.67% | Not specified |
| 20 | DeepSeek V3.1 | 89.67% | Not specified |
| 23 | gpt-oss-20b | 89.33% | high |
| 23 | Grok 4.1 Fast | 89.33% | Not specified |
| 23 | Ring-1T | 89.33% | Not specified |
| 26 | Nova 2.0 Pro Preview | 89.00% | medium |
| 27 | o3 | 88.33% | Not specified |
| 27 | Qwen3 VL 235B A22B (Reasoning) | 88.33% | Not specified |
| 29 | INTELLECT-3 | 88.00% | Not specified |
| 29 | Claude 4.5 Sonnet | 88.00% | Not specified |
| 29 | Apriel-v1.6-15B-Thinker | 88.00% | Not specified |
| 32 | Gemini 2.5 Pro | 87.67% | Not specified |
| 33 | Apriel-v1.5-15B-Thinker | 87.50% | Not specified |
| 34 | GLM-4.6 | 86.00% | Not specified |
| 35 | ERNIE 5.0 Thinking Preview | 85.00% | Not specified |
| 36 | Seed-OSS-36B-Instruct | 84.67% | Not specified |
| 36 | Qwen3 VL 32B (Reasoning) | 84.67% | Not specified |
| 36 | Grok 3 mini Reasoning | 84.67% | high |
| 39 | Qwen3 Next 80B A3B (Reasoning) | 84.33% | Not specified |
| 40 | GPT-5 nano | 83.67% | high |
| 40 | Ring-flash-2.0 | 83.67% | Not specified |
| 40 | Claude 4.5 Haiku | 83.67% | Not specified |
| 43 | MiniMax-M2.1 | 82.67% | Not specified |
| 43 | Qwen3 4B 2507 (Reasoning) | 82.67% | Not specified |
| 45 | Qwen3 Max Thinking (Preview) | 82.33% | Not specified |
| 45 | Qwen3 VL 30B A3B (Reasoning) | 82.33% | Not specified |
| 47 | Magistral Medium 1.2 | 82.00% | Not specified |
| 48 | GLM-4.5-Air | 80.67% | Not specified |
| 48 | Qwen3 Max | 80.67% | Not specified |
| 50 | Motif-2-12.7B-Reasoning | 80.33% | Not specified |
| 50 | Magistral Small 1.2 | 80.33% | Not specified |
| 50 | Claude 4.1 Opus | 80.33% | Not specified |
| 53 | Falcon-H1R-7B | 80.00% | Not specified |
| 54 | o1-preview | 79.67% | Not specified |
| 55 | Doubao Seed Code | 79.33% | Not specified |
| 56 | Mi:dm K 2.5 Pro Preview | 78.67% | Not specified |
| 57 | K2-V2 | 78.33% | high |
| 57 | MiniMax-M2 | 78.33% | Not specified |
| 59 | Olmo 3.1 32B Think | 77.33% | Not specified |
| 60 | Mi:dm K 2.5 Pro | 76.67% | Not specified |
| 61 | DeepSeek R1 0528 (May '25) | 76.00% | Not specified |
| 62 | Qwen3 Max (Preview) | 75.00% | Not specified |
| 62 | NVIDIA Nemotron Nano 12B v2 VL | 75.00% | Not specified |
| 64 | Claude 4 Sonnet | 74.33% | Not specified |
| 65 | Qwen3 Omni 30B A3B (Reasoning) | 74.00% | Not specified |
| 66 | GLM-4.5 | 73.67% | Not specified |
| 66 | Olmo 3 32B Think | 73.67% | Not specified |
| 68 | Gemini 2.5 Flash | 73.33% | Not specified |
| 69 | Qwen3 32B | 73.00% | Not specified |
| 70 | Cogito v2.1 | 72.67% | Not specified |
| 71 | Qwen3 30B A3B | 72.33% | Not specified |
| 71 | Qwen3 VL 30B A3B Instruct | 72.33% | Not specified |
| 73 | Qwen3 235B A22B 2507 | 71.67% | Not specified |
| 74 | Ling-1T | 71.33% | Not specified |
| 75 | Olmo 3 7B Think | 70.67% | Not specified |
| 75 | Qwen3 VL 235B A22B Instruct | 70.67% | Not specified |
| 77 | Hermes 4 - Llama-3.1 405B | 69.67% | Not specified |
| 78 | Qwen3 VL 32B Instruct | 68.33% | Not specified |
| 79 | DeepSeek R1 (Jan '25) | 68.00% | Not specified |
| 80 | MiMo-V2-Flash | 67.67% | Not specified |
| 81 | Qwen3 Next 80B A3B Instruct | 66.33% | Not specified |
| 81 | Qwen3 30B A3B 2507 | 66.33% | Not specified |
| 83 | Ling-flash-2.0 | 65.33% | Not specified |
| 84 | DeepSeek R1 0528 Qwen3 8B | 63.67% | Not specified |
| 84 | Llama 3.1 Nemotron Ultra 253B v1 | 63.67% | Not specified |
| 86 | DeepSeek R1 Distill Qwen 32B | 63.00% | Not specified |
| 87 | NVIDIA Nemotron Nano 9B V2 | 62.33% | Not specified |
| 88 | Solar Pro 2 | 61.33% | high |
| 89 | MiniMax M1 80k | 61.00% | Not specified |
| 90 | HyperCLOVA X SEED Think (32B) | 59.00% | Not specified |
| 90 | DeepSeek V3.2 | 59.00% | Not specified |
| 92 | Grok 3 | 58.00% | Not specified |
| 93 | DeepSeek V3.2 Exp | 57.67% | Not specified |
| 94 | Kimi K2 0905 | 57.33% | Not specified |
| 95 | Kimi K2 | 57.00% | Not specified |
| 96 | Gemini 2.5 Flash Preview (Sep '25) | 56.67% | Not specified |
| 97 | Claude 3.7 Sonnet | 56.33% | Not specified |
| 97 | Qwen3 30B A3B 2507 (Reasoning) | 56.33% | Not specified |
| 99 | Qwen3 14B | 55.67% | Not specified |
| 99 | Gemini 3 Flash Preview | 55.67% | Not specified |
| 99 | DeepSeek R1 Distill Qwen 14B | 55.67% | Not specified |
| 102 | Llama 3.3 Nemotron Super 49B v1 | 54.67% | Not specified |
| 103 | DeepSeek R1 Distill Llama 70B | 53.67% | Not specified |
| 104 | Qwen3 Omni 30B A3B Instruct | 52.33% | Not specified |
| 104 | Qwen3 4B 2507 | 52.33% | Not specified |
| 106 | Llama 3.1 Nemotron Nano 4B v1.1 | 50.00% | Not specified |
| 107 | Ling-mini-2.0 | 49.33% | Not specified |
| 108 | GPT-5 (ChatGPT) | 48.33% | Not specified |
| 109 | Gemini 2.5 Flash-Lite Preview (Sep '25) | 46.67% | Not specified |
| 110 | GPT-4.1 mini | 46.33% | Not specified |
| 111 | Grok Code Fast 1 | 43.33% | Not specified |
| 112 | DeepSeek R1 Distill Llama 8B | 41.33% | Not specified |
| 112 | Magistral Small 1 | 41.33% | Not specified |
| 112 | Grok 4 Fast | 41.33% | Not specified |
| 112 | Olmo 3 7B Instruct | 41.33% | Not specified |
| 112 | ERNIE 4.5 300B A47B | 41.33% | Not specified |
| 117 | DeepSeek V3 0324 | 41.00% | Not specified |
| 118 | Magistral Medium 1 | 40.33% | Not specified |
| 119 | Qwen3 Coder 480B A35B Instruct | 39.33% | Not specified |
| 119 | EXAONE 4.0 32B | 39.33% | Not specified |
| 121 | Mistral Medium 3.1 | 38.33% | Not specified |
| 122 | Mistral Large 3 | 38.00% | Not specified |
| 123 | Qwen3 VL 4B Instruct | 37.00% | Not specified |
| 124 | Devstral 2 | 36.67% | Not specified |
| 125 | Claude 4 Opus | 36.33% | Not specified |
| 125 | Kimi Linear 48B A3B Instruct | 36.33% | Not specified |
| 127 | Gemini 2.5 Flash-Lite | 35.33% | Not specified |
| 128 | GPT-4.1 | 34.67% | Not specified |
| 129 | Devstral Small 2 | 34.33% | Not specified |
| 130 | Reka Flash 3 | 33.67% | Not specified |
| 131 | Ministral 3 8B | 31.67% | Not specified |
| 132 | Qwen3 VL 8B (Reasoning) | 30.67% | Not specified |
| 133 | Mistral Medium 3 | 30.33% | Not specified |
| 134 | Ministral 3 14B | 30.00% | Not specified |
| 135 | Devstral Small (Jul '25) | 29.33% | Not specified |
| 136 | QwQ 32B | 29.00% | Not specified |
| 136 | Qwen3 Coder 30B A3B Instruct | 29.00% | Not specified |
| 138 | Qwen3 VL 8B Instruct | 27.33% | Not specified |
| 139 | Mistral Small 3.2 | 27.00% | Not specified |
| 140 | GLM-4.6V | 26.33% | Not specified |
| 141 | DeepSeek V3 (Dec '24) | 26.00% | Not specified |
| 142 | GPT-4o (March 2025, chatgpt-4o-latest) | 25.67% | Not specified |
| 142 | Qwen3 VL 4B (Reasoning) | 25.67% | Not specified |
| 144 | LFM2 8B A1B | 25.33% | Not specified |
| 145 | GPT-4.1 nano | 24.00% | Not specified |
| 145 | Exaone 4.0 1.2B | 24.00% | Not specified |
| 147 | Qwen3 235B A22B | 23.67% | Not specified |
| 148 | Qwen3 4B | 22.33% | Not specified |
| 149 | DeepSeek R1 Distill Qwen 1.5B | 22.00% | Not specified |
| 149 | Ministral 3 3B | 22.00% | Not specified |
| 151 | Gemini 2.0 Flash (Feb '25) | 21.67% | Not specified |
| 152 | Gemma 3 27B Instruct | 20.67% | Not specified |
| 153 | Llama 4 Maverick | 19.33% | Not specified |
| 154 | Qwen3 8B | 19.00% | Not specified |
| 155 | Gemma 3 12B Instruct | 18.33% | Not specified |
| 156 | Phi-4 | 18.00% | Not specified |
| 157 | Nova Premier | 17.33% | Not specified |
| 158 | GLM-4.5V | 15.33% | Not specified |
| 159 | GPT-4o mini | 14.67% | Not specified |
| 160 | Gemma 3n E4B Instruct | 14.33% | Not specified |
| 161 | Mistral Large 2 (Nov '24) | 14.00% | Not specified |
| 161 | Llama 4 Scout | 14.00% | Not specified |
| 161 | Qwen2.5 Instruct 72B | 14.00% | Not specified |
| 164 | MiniMax M1 40k | 13.67% | Not specified |
| 164 | Granite 4.0 H Small | 13.67% | Not specified |
| 166 | NVIDIA Nemotron 3 Nano 30B A3B | 13.33% | Not specified |
| 167 | Command A | 13.00% | Not specified |
| 168 | Gemma 3 4B Instruct | 12.67% | Not specified |
| 169 | Hermes 4 - Llama-3.1 70B | 11.33% | Not specified |
| 170 | Llama 3.1 Nemotron Instruct 70B | 11.00% | Not specified |
| 171 | Jamba Reasoning 3B | 10.67% | Not specified |
| 172 | Qwen3 0.6B | 10.33% | Not specified |
| 172 | Gemma 3n E2B Instruct | 10.33% | Not specified |
| 174 | LFM2 2.6B | 8.33% | Not specified |
| 175 | Llama Nemotron Super 49B v1.5 | 8.00% | Not specified |
| 176 | Llama 3.3 Instruct 70B | 7.67% | Not specified |
| 177 | Qwen3 1.7B | 7.33% | Not specified |
| 178 | Nova Pro | 7.00% | Not specified |
| 178 | Nova Lite | 7.00% | Not specified |
| 180 | Granite 3.3 8B | 6.67% | Not specified |
| 180 | Phi-4 Mini Instruct | 6.67% | Not specified |
| 182 | Granite 4.0 1B | 6.33% | Not specified |
| 182 | Granite 4.0 H 1B | 6.33% | Not specified |
| 184 | Granite 4.0 Micro | 6.00% | Not specified |
| 184 | Nova Micro | 6.00% | Not specified |
| 184 | GPT-4o (Nov '24) | 6.00% | Not specified |
| 187 | Devstral Medium | 4.67% | Not specified |
| 188 | Llama 3.1 Instruct 8B | 4.33% | Not specified |
| 188 | Mistral Small 3 | 4.33% | Not specified |
| 190 | Llama 3.1 Instruct 70B | 4.00% | Not specified |
| 191 | Mistral Small 3.1 | 3.67% | Not specified |
| 192 | LFM2 1.2B | 3.33% | Not specified |
| 192 | OLMo 2 32B | 3.33% | Not specified |
| 192 | Llama 3.2 Instruct 3B | 3.33% | Not specified |
| 192 | Gemma 3 1B Instruct | 3.33% | Not specified |
| 196 | Llama 3.1 Instruct 405B | 3.00% | Not specified |
| 197 | Gemma 3 270M | 2.33% | Not specified |
| 197 | Jamba 1.7 Large | 2.33% | Not specified |
| 197 | Pixtral Large | 2.33% | Not specified |
| 200 | Llama 3.2 Instruct 11B (Vision) | 1.67% | Not specified |
| 201 | Granite 4.0 H 350M | 1.33% | Not specified |
| 202 | OLMo 2 7B | 0.67% | Not specified |
| 203 | Jamba 1.7 Mini | 0.33% | Not specified |
| 203 | Phi-3 Mini Instruct 3.8B | 0.33% | Not specified |
| 205 | Mistral Large 2 (Jul '24) | 0.00% | Not specified |
| 205 | Granite 4.0 350M | 0.00% | Not specified |
| 205 | Llama 3.2 Instruct 1B | 0.00% | Not specified |
| 205 | Molmo 7B-D | 0.00% | Not specified |
No published result tracked
46 of the 46 models on this site have no published number on this benchmark, so their cells in the home table stay blank rather than reading as zero.
Step 5 PreviewGPT-6 LunaGPT-5.6 TerraGLM 5.3 FlashQwen3.8 2.4T A95BQwen3.8-Flash-NextMiMo-V2.6-FlashQwen3.8 Max (0902)Claude Fable 5Grok 4.6Claude Opus 4.8Claude Opus 4.7Gemini 3.7 FlashMuse Spark 1.2GPT-5.4Grok 4.5GPT-5.5DeepSeek V4.1 Flash 1017GLM-LegalClaude Haiku 5.5BeamGemini 4 ArgonClaude Sonnet 5.5Claude Opus 5.5Claude Fable 5.1Claude Opus 5Claude Sonnet 5GPT-6 AstraGPT-6 SolGPT-6.1 SolGPT-5.6 SolGPT-5.6 LunaGemini 3.8 FlashGemini 3.1 ProGrok 4.7DeepSeek V4.1 FlashDeepSeek V4 ProKimi K3Qwen3.8 MaxGLM 5.3Muse Spark 1.3 MaxMiniMax M3Hunyuan Hy4 PreviewMiMo V2.6 ProInklingMistral Medium 3.5
Read this with
Benchmarks in the same category, which usually disagree with this one on purpose.
Sources
Every page used for this benchmark, read on 17 October 2026.
- Benchmark site or the page that publishes itartificialanalysis.ai/evaluations/aime-2025
- Independent resultsartificialanalysis.ai/evaluations/aime-2025