llmrun
Math

AIME 2024/2025 Leaderboard

AIME (American Invitational Mathematics Examination) is a prestigious high-school competition of hard, integer-answer problems. It's a widely-cited yardstick for multi-step mathematical reasoning, here on the 2024–2025 papers.

Source: epoch34 open models ranked+121 proprietaryData through Jul 2026

Open modelsAll models

All models ranked on AIME 2024/2025

Proprietary / closed models are shown dimmed — you can't run them locally, but they show where the open field stands.

#ModelScore
1GPT 5.5 Pre Release (xhigh) · proprietary
100.0%
2GPT 5.5 Pro Pre Release (xhigh) · proprietary
100.0%
3GPT 5.6 Sol Max · proprietary
100.0%
4Claude Fable 5 Max · proprietary
99.7%
5GPT 5.6 Terra Max · proprietary
99.7%
6Claude Opus 4.8 Max · proprietary
98.3%
7GPT 5.6 Luna Max · proprietary
98.3%
8Claude Opus 4.7 (xhigh) · proprietary
97.8%
9Grok 4.5 (high) · proprietary
97.8%
10Kimi K3 Max · proprietary
97.2%
11DeepSeek V4 Pro · 861.6B
96.7%
12Kimi K2.7 Code · 1058.6B
96.4%
13GPT 5.2 (Dec 11, 2025, high) · proprietary
96.1%
14Kimi K2.6 · 1058.6B
96.1%
15GPT 5.2 (Dec 11, 2025, xhigh) · proprietary
96.1%
16Gemini 3.1 Pro Preview · proprietary
95.6%
17Gemini 3.5 Flash (high) · proprietary
95.6%
18GPT 5.4 (Mar 05, 2026, xhigh) · proprietary
95.3%
19Qwen3.7 Max · proprietary
95.0%
20Claude Sonnet 5 (xhigh) · proprietary
94.7%
21Claude Opus 4.6 (64K) · proprietary
94.4%
22GPT 5.2 (Dec 11, 2025, medium) · proprietary
93.9%
23Grok 4.3 (high) · proprietary
93.3%
24Claude Opus 4.6 (32K) · proprietary
93.1%
25Gemini 3 Flash Preview · proprietary
92.8%
26GLM 5.1 · 753.9B
92.2%
27Grok 4.20 0309 Reasoning · proprietary
92.2%
28Kimi K2.5 · 1058.6B
92.2%
29Gemini 3 Pro Preview · proprietary
91.4%
30GPT 5 (Aug 07, 2025, high) · proprietary
91.4%
31Qwen3.6 Max Preview · proprietary
91.1%
32Qwen3.6 Plus · proprietary
90.6%
33Muse Spark · proprietary
88.9%
34GPT OSS 120B · 120.4B
88.9%
35GPT 5.1 (Nov 13, 2025, high) · proprietary
88.6%
36DeepSeek Reasoner · proprietary
87.8%
37GPT 5.4 Nano (Mar 17, 2026, high) · proprietary
87.8%
38GPT 5 (Aug 07, 2025, medium) · proprietary
87.2%
39GPT 5.4 Mini (Mar 17, 2026, high) · proprietary
87.2%
40GPT 5 Mini (Aug 07, 2025, high) · proprietary
86.7%
41Qwen3 235B A22B Thinking 2507 · 235.1B
86.7%
42GLM 5.2 Max · proprietary
86.4%
43Claude Opus 4.5 (Nov 01, 2025, 32K) · proprietary
86.1%
44Qwen3.6 Flash · proprietary
86.1%
45Claude Sonnet 4.6 (32K) · proprietary
85.8%
46GPT 5.1 (Nov 13, 2025, medium) · proprietary
85.6%
47Qwen3.5 Flash · proprietary
85.6%
48Qwen3.5 Plus · proprietary
85.0%
49Gemini 2.5 Pro · proprietary
84.2%
50Grok 4 (Jul 09) · proprietary
84.0%
51O3 (Apr 16, 2025, high) · proprietary
83.9%
52GLM 4.7 · 358.3B
83.3%
53Kimi K2 Thinking · 1058.1B
83.1%
54Claude Opus 4.5 (Nov 01, 2025, 16K) · proprietary
81.7%
55O4 Mini (Apr 16, 2025, high) · proprietary
81.7%
56GPT 5 Nano (Aug 07, 2025, high) · proprietary
81.1%
57GLM 5 · 753.9B
80.0%
58GPT 5.2 (Dec 11, 2025, low) · proprietary
78.9%
59GPT 5 Mini (Aug 07, 2025, medium) · proprietary
78.3%
60Claude Sonnet 4.5 (Sep 29, 2025, 32K) · proprietary
77.8%
61Claude Sonnet 4.5 (Sep 29, 2025, 59K) · proprietary
77.8%
62Grok 3 Mini Beta (high) · proprietary
77.8%
63O3 Mini (Jan 31, 2025, high) · proprietary
76.9%
64GPT 5 Nano (Aug 07, 2025, medium) · proprietary
74.2%
65O1 (Dec 17, 2024, medium) · proprietary
73.3%
66Qwen3 Max (Sep 23, 2025) · proprietary
73.3%
67Gemini 2.5 Flash Preview (Apr 17) · proprietary
73.1%
68Claude Sonnet 4 (May 14, 2025, 32K) · proprietary
71.1%
69Claude Sonnet 4.5 (Sep 29, 2025, 16K) · proprietary
71.1%
70Gemini 2.5 Flash Preview (May 20) · proprietary
70.8%
71Claude Opus 4.1 (Aug 05, 2025, 27K) · proprietary
68.9%
72Claude Sonnet 4 (May 14, 2025, 59K) · proprietary
68.9%
73Claude Haiku 4.5 (Oct 01, 2025, 32K) · proprietary
66.7%
74DeepSeek R1 0528 · 684.5B
66.4%
75Claude Opus 4 (May 14, 2025, 27K) · proprietary
64.4%
76Claude Opus 4.1 (Aug 05, 2025, 16K) · proprietary
64.4%
77GPT 5.1 (Nov 13, 2025, low) · proprietary
63.9%
78O3 Mini (Jan 31, 2025, medium) · proprietary
63.9%
79Grok 3 Mini Beta (low) · proprietary
62.2%
80Claude Opus 4 (May 14, 2025, 16K) · proprietary
60.0%
81Claude 3.7 Sonnet (Feb 19, 2025, 64K) · proprietary
57.8%
82Gemini 2.0 Flash Thinking Exp (Jan 21) · proprietary
57.8%
83Grok 3 Beta · proprietary
55.6%
84Claude 3.7 Sonnet (Feb 19, 2025, 32K) · proprietary
53.3%
85Claude Sonnet 4 (May 14, 2025, 16K) · proprietary
53.3%
86DeepSeek R1 · 684.5B
53.3%
87DeepSeek R1 Distill Llama 70B · 70.6B
51.4%
88Claude Opus 4.5 (Nov 01, 2025) · proprietary
48.1%
89O1 Mini (Sep 12, 2024, high) · proprietary
46.9%
90Claude 3.7 Sonnet (Feb 19, 2025, 16K) · proprietary
46.7%
91GPT 4.1 Mini (Apr 14, 2025) · proprietary
44.7%
92O1 Mini (Sep 12, 2024, medium) · proprietary
44.7%
93Claude Opus 4 (May 14, 2025) · proprietary
42.2%
94Claude Opus 4.1 (Aug 05, 2025) · proprietary
40.0%
95GPT 4.1 (Apr 14, 2025) · proprietary
38.3%
96DeepSeek v3 0324 · 684.5B
37.8%
97GPT 4.5 Preview (Feb 27, 2025) · proprietary
37.8%
98Claude Haiku 4.5 (Oct 01, 2025) · proprietary
35.8%
99Claude Sonnet 4.5 (Sep 29, 2025) · proprietary
35.6%
100Mistral Medium 2505 · proprietary
32.2%
101Gemini 2.0 Flash 001 · proprietary
31.1%
102O1 Preview (Sep 12, 2024) · proprietary
31.1%
103Magistral Small 2506 · 23.6B
30.0%
104Claude Sonnet 4 (May 14, 2025) · proprietary
28.9%
105GPT 4.1 Nano (Apr 14, 2025) · proprietary
28.9%
106Gemini 1.5 Pro 002 · proprietary
23.1%
107Claude 3.7 Sonnet (Feb 19, 2025) · proprietary
21.9%
108Llama 4 Maverick 17B 128E Instruct · 401.6B
20.6%
109Gemma 3 27B IT · 27.4B
19.7%
110Qwen Plus (Jan 25, 2025) · proprietary
17.8%
111Gemini 1.5 Flash 002 · proprietary
16.3%
112Qwen Max (Jan 25, 2025) · proprietary
16.1%
113DeepSeek v3 · 684.5B
15.8%
114Phi 4 · 14.7B
13.8%
115Grok 2 (Dec 12) · proprietary
11.5%
116Llama 3.1 405B Instruct · 405.9B
9.7%
117Claude 3.5 Sonnet (Oct 22, 2024) · proprietary
8.5%
118Mistral Large 2407 · proprietary
8.5%
119Qwen2.5 72B Instruct · 72.7B
8.1%
120Llama 4 Scout 17B 16E Instruct · 108.6B
7.8%
121Mistral Large 2411 · proprietary
7.8%
122Qwen2.5 32B Instruct · 32.8B
7.4%
123GPT 4o Mini (Jul 18, 2024) · proprietary
6.9%
124Gemini 1.5 Pro 001 · proprietary
6.8%
125GPT 4 Turbo (Apr 09, 2024) · proprietary
6.7%
126Claude 3.5 Sonnet (Jun 20, 2024) · proprietary
6.5%
127GPT 4o (Aug 06, 2024) · proprietary
6.4%
128GPT 4o (May 13, 2024) · proprietary
6.3%
129GPT 4o (Nov 20, 2024) · proprietary
6.3%
130Qwen Turbo (Nov 01, 2024) · proprietary
6.1%
131Mistral Small 2503 · proprietary
5.8%
132Mistral Small 2501 · proprietary
5.3%
133Llama 3.3 70B Instruct · 70.6B
5.1%
134Claude 3 Opus (Feb 29, 2024) · proprietary
4.7%
135Gemini 1.5 Flash 8B 001 · proprietary
4.6%
136Llama 3.1 Tulu 3 70B DPO · 70.6B
4.4%
137Claude 3.5 Haiku (Oct 22, 2024) · proprietary
4.3%
138Meta Llama 3 70B Instruct · 70.6B
4.3%
139Gemini 1.5 Flash 001 · proprietary
3.9%
140Llama 3.1 70B Instruct · 70.6B
3.6%
141Llama 3.2 90B Vision Instruct · 88.6B
2.6%
142Claude 2.0 · proprietary
2.5%
143Claude 3 Sonnet (Feb 29, 2024) · proprietary
2.5%
144Hermes 2 Theta Llama 3 70B · 70.6B
2.5%
145Llama 3.1 8B Instruct · 8.0B
2.5%
146Claude 2.1 · proprietary
1.9%
147Mistral Large 2402 · proprietary
1.9%
148Claude 3 Haiku (Mar 07, 2024) · proprietary
1.8%
149Gemma 2 27B IT · 27.2B
1.4%
150Gemini 1.0 Pro 001 · proprietary
1.1%
151GPT 4 (Jun 13) · proprietary
1.1%
152Meta Llama 3 8B Instruct · 8.0B
0.8%
153Gemma 2 9B IT · 9.2B
0.6%
154GPT 4 (Mar 14) · proprietary
0.6%
155Llama 2 70B Chat HF · 69.0B
0.0%

Score vs model size

Which models give the most quality for their size — the ones worth running locally.

10B100B1Tmodel size (log scale) →96.7%0.0%Kimi K2.7 Code · 1.1T · 96.4%Kimi K2.6 · 1.1T · 96.1%Kimi K2.5 · 1.1T · 92.2%Qwen3 235B A22B Thinking 2507 · 235B · 86.7%GLM 4.7 · 358B · 83.3%Kimi K2 Thinking · 1.1T · 83.1%GLM 5 · 754B · 80.0%DeepSeek R1 0528 · 685B · 66.4%DeepSeek R1 · 685B · 53.3%DeepSeek v3 0324 · 685B · 37.8%Llama 4 Maverick 17B 128E Instruct · 402B · 20.6%Gemma 3 27B IT · 27B · 19.7%DeepSeek v3 · 685B · 15.8%Llama 3.1 405B Instruct · 406B · 9.7%Qwen2.5 72B Instruct · 73B · 8.1%Llama 4 Scout 17B 16E Instruct · 109B · 7.8%Qwen2.5 32B Instruct · 33B · 7.4%Llama 3.3 70B Instruct · 71B · 5.1%Llama 3.1 Tulu 3 70B DPO · 71B · 4.4%Meta Llama 3 70B Instruct · 71B · 4.3%Llama 3.1 70B Instruct · 71B · 3.6%Llama 3.2 90B Vision Instruct · 89B · 2.6%Hermes 2 Theta Llama 3 70B · 71B · 2.5%Gemma 2 27B IT · 27B · 1.4%Meta Llama 3 8B Instruct · 8B · 0.8%Gemma 2 9B IT · 9B · 0.6%Llama 2 70B Chat HF · 69B · 0.0%Llama 3.1 8B Instruct · 8B · 2.5%Llama 3.1 8B InstructPhi 4 · 15B · 13.8%Phi 4Magistral Small 2506 · 24B · 30.0%Magistral Small 2506DeepSeek R1 Distill Llama 70B · 71B · 51.4%DeepSeek R1 Distill L…GPT OSS 120B · 120B · 88.9%GPT OSS 120BGLM 5.1 · 754B · 92.2%GLM 5.1DeepSeek V4 Pro · 862B · 96.7%DeepSeek V4 Pro
Each dot is a model. Up = higher score, left = smaller (easier to run locally). The dashed line marks the efficiency frontier — the best score you can get at each size or smaller.
  • Llama 3.1 8B Instruct, 8B, score 2.5% — on the efficiency frontier (best score at its size or smaller).
  • Phi 4, 15B, score 13.8% — on the efficiency frontier (best score at its size or smaller).
  • Magistral Small 2506, 24B, score 30.0% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek R1 Distill Llama 70B, 71B, score 51.4% — on the efficiency frontier (best score at its size or smaller).
  • GPT OSS 120B, 120B, score 88.9% — on the efficiency frontier (best score at its size or smaller).
  • GLM 5.1, 754B, score 92.2% — on the efficiency frontier (best score at its size or smaller).
  • DeepSeek V4 Pro, 862B, score 96.7% — on the efficiency frontier (best score at its size or smaller).

AIME 2024/2025: frequently asked questions

What is the best open LLM on AIME 2024/2025?
DeepSeek V4 Pro is the top open model on AIME 2024/2025, scoring 96.7%. Among all models tested — including proprietary ones — it ranks #11. The top model overall is GPT 5.5 Pro Pre Release (xhigh) (OpenAI) at 100.0%.
What's the best AIME 2024/2025 model you can run on a 24 GB GPU?
Magistral Small 2506 is the highest-scoring open model that fits in 24 GB at 4-bit quantization (about 13 GB), scoring 30.0% on AIME 2024/2025.
What's the best AIME 2024/2025 model you can run on a 12 GB GPU?
Phi 4 is the highest-scoring open model that fits in 12 GB at 4-bit quantization (about 8 GB), scoring 13.8% on AIME 2024/2025.
Can open models match proprietary models on AIME 2024/2025?
Not quite on AIME 2024/2025: the strongest proprietary model (GPT 5.5 Pro Pre Release (xhigh)) scores 100.0%, ahead of the best open model (DeepSeek V4 Pro) at 96.7% — but you can run the open one yourself.

Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.