Benchmark Almanac

AIME 2024

gpt-oss-120b 96.6% leads since Aug 2025 Fading

Models solve the 30 problems of the 2024 AIME I and II. Answers are integers from 0 to 999, scored by exact match.

The problems and solutions were published in February 2024, so later training data can contain them.

Made by Mathematical Association of America Released Feb 2024 Competition problems with answers All tasks public Last in a launch post Jan 2026 , dropped after it saturated Last model added Feb 2026
Level
Competition The strongest students, in olympiad-style contests
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Short answer A number, word or short string matched to a key.
Dataset artofproblemsolving.com
87
Models scored in lab cards
70
Launch posts list it, from 29 labs
57
System cards report it, from 27 labs

gpt-oss-120b leads with 96.6% among models released by Feb 2026

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

The newest model with a score here came out in Feb 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

87 models. Scroll sideways for the rest.

  1. 96.6 gpt-oss-120b
  2. 96.0 gpt-oss-20b
  3. 95.8 Grok-3 mini
  4. 93.4 o4-mini
  5. 93.3 Grok 3
  6. 93.3 LongCat-Flash-Thinking
  7. 93.1 DeepSeek-V3.1
  8. 93.0 o3-pro
  9. 92.0 Gemini 2.5 Pro
  10. 91.8 Magistral Medium 1.2
  11. 91.7 seed-oss-36b-instruct
  12. 91.7 Solar Open 100B
  13. 91.6 o3
  14. 91.4 DeepSeek-R1
  15. 91.0 GLM-4.5
  16. 90.9 STEP3-VL-10B
  17. 90.8 INTELLECT-3
  18. 90.0 Magistral Medium 1.0
  19. 90.0 Seed1.6-Thinking
  20. 89.8 Ministral 3 14B
  21. 89.4 GLM-4.5-Air
  22. 89.0 qwen3-coder-next
  23. 88.1 Falcon H1R 7B
  24. 88.0 Gemini 2.5 Flash
  25. 87.5 Llama-3.3-Nemotron-Super-49B-v1.5
  26. 87.3 o3-mini
  27. 87.3 Hunyuan-A13B-Instruct
  28. 86.7 Seed-Thinking-v1.5
  29. 86.1 Magistral Small 1.2
  30. 86.0 MiniMax-M1-80k
  31. 85.7 Qwen3-235B-A22B
  32. 84.7 InternVL3.5-241B-A28B
  33. 83.4 Baichuan-M2-32B
  34. 81.9 nvidia-nemotron-nano-9b-v2
  35. 81.9 Hermes 4 405B
  36. 81.4 Qwen3-32B
  37. 81.3 Phi-4-Reasoning-plus
  38. 80.8 GLM-Z1-32B-0414
  39. 80.8 Llama-3.1-Nemotron-Ultra-253B-v1
  40. 80.6 Olmo 3.1 32B Think
  41. 80.4 Qwen3-30B-A3B
  42. 80.0 Claude 3.7 Sonnet
  43. 79.3 Qwen3-14B
  44. 79.2 o1
  45. 78.8 INTELLECT-2
  46. 78.2 Hunyuan-T1
  47. 77.5 Kimi k1.5
  48. 76.8 Olmo 3 32B Think
  49. 76.5 Solar Pro 2
  50. 76.0 Qwen3-8B
  51. 76.0 Code World Model (CWM)
  52. 75.3 Phi-4-reasoning
  53. 73.8 Qwen3-4B
  54. 73.3 Gemini 2.0 Flash Thinking
  55. 72.9 LongCat-Flash-Omni
  56. 72.6 DeepSeek-R1-Distill-Qwen-32B
  57. 72.1 EXAONE Deep 32B
  58. 72.0 Magistral Medium 1.1
  59. 71.9 Hermes 4.3 36B
  60. 70.7 Magistral Small 1.0
  61. 70.5 Magistral Small 1.1
  62. 70.4 LongCat-Flash-Chat
  63. 70.0 DeepSeek-R1-Distill-Llama-70B
  64. 70.0 Reka Flash 3
  65. 69.7 DeepSeek-R1-Distill-Qwen-14B
  66. 69.6 Kimi K2
  67. 68.2 MiMo-7B-RL
  68. 67.5 Llama-3.3-Nemotron-Super-49B-v1
  69. 59.4 DeepSeek-V3
  70. 57.5 Phi-4-mini-reasoning
  71. 56.7 o1-preview
  72. 54.8 ERNIE-4.5-300B-A47B
  73. 52.3 Phi-4-mini-flash-reasoning
  74. 49.6 GPT-4.1 mini
  75. 48.3 Qwen3-1.7B
  76. 43.3 Hunyuan-TurboS
  77. 36.7 GPT-4.5
  78. 30.0 Mercury
  79. 29.4 GPT-4.1 nano
  80. 28.9 DeepSeek-R1-Distill-Qwen-1.5B
  81. 27.6 Claude 3.5 Sonnet (October 2024)
  82. 23.8 Falcon-H1-34B
  83. 23.3 Cohere Command A
  84. 20.0 InternLM3-8B-Instruct
  85. 10.0 Phi-4 Mini
  86. 8.12 Granite 3.3 8B Instruct
  87. 5.30 Claude 3.5 Haiku
  1. 1 gpt-oss-120b 96.6%
  2. 2 gpt-oss-20b 96%
  3. 3 Grok-3 mini 95.8%
  4. 4 o4-mini 93.4%
  5. 5 Grok 3 93.3%
  6. 6 LongCat-Flash-Thinking 93.3%
  7. 7 DeepSeek-V3.1 93.1%
  8. 8 o3-pro 93%
  9. 9 Gemini 2.5 Pro Mar 2025 92%
  10. 10 Magistral Medium 1.2 91.8%
  11. 11 seed-oss-36b-instruct 91.7%
  12. 12 Solar Open 100B 91.7%
  13. 13 o3 91.6%
  14. 14 DeepSeek-R1 91.4%
  15. 15 GLM-4.5 91%
  16. 16 STEP3-VL-10B 90.9%
  17. 17 INTELLECT-3 90.8%
  18. 18 Magistral Medium 1.0 90%
  19. 19 Seed1.6-Thinking 90%
  20. 20 Ministral 3 14B 89.8%
  21. 21 GLM-4.5-Air 89.4%
  22. 22 qwen3-coder-next 89%
  23. 23 Falcon H1R 7B 88.1%
  24. 24 Gemini 2.5 Flash Apr 2025 88%
  25. 25 Llama-3.3-Nemotron-Super-49B-v1.5 87.5%
  26. 26 o3-mini 87.3%
  27. 27 Hunyuan-A13B-Instruct 87.3%
  28. 28 Seed-Thinking-v1.5 86.7%
  29. 29 Magistral Small 1.2 86.1%
  30. 30 MiniMax-M1-80k 86%
  31. 31 Qwen3-235B-A22B 85.7%
  32. 32 InternVL3.5-241B-A28B 84.7%
  33. 33 Baichuan-M2-32B 83.4%
  34. 34 nvidia-nemotron-nano-9b-v2 81.9%
  35. 35 Hermes 4 405B 81.9%
  36. 36 Qwen3-32B 81.4%
  37. 37 Phi-4-Reasoning-plus 81.3%
  38. 38 GLM-Z1-32B-0414 80.8%
  39. 39 Llama-3.1-Nemotron-Ultra-253B-v1 80.8%
  40. 40 Olmo 3.1 32B Think 80.6%
  41. 41 Qwen3-30B-A3B 80.4%
  42. 42 Claude 3.7 Sonnet 80%
  43. 43 Qwen3-14B 79.3%
  44. 44 o1 79.2%
  45. 45 INTELLECT-2 78.8%
  46. 46 Hunyuan-T1 78.2%
  47. 47 Kimi k1.5 77.5%
  48. 48 Olmo 3 32B Think 76.8%
  49. 49 Solar Pro 2 76.5%
  50. 50 Qwen3-8B 76%
  51. 51 Code World Model (CWM) 76%
  52. 52 Phi-4-reasoning 75.3%
  53. 53 Qwen3-4B 73.8%
  54. 54 Gemini 2.0 Flash Thinking Jan 2025 73.3%
  55. 55 LongCat-Flash-Omni 72.9%
  56. 56 DeepSeek-R1-Distill-Qwen-32B 72.6%
  57. 57 EXAONE Deep 32B 72.1%
  58. 58 Magistral Medium 1.1 72%
  59. 59 Hermes 4.3 36B 71.9%
  60. 60 Magistral Small 1.0 70.7%
  61. 61 Magistral Small 1.1 70.5%
  62. 62 LongCat-Flash-Chat 70.4%
  63. 63 DeepSeek-R1-Distill-Llama-70B 70%
  64. 64 Reka Flash 3 70%
  65. 65 DeepSeek-R1-Distill-Qwen-14B 69.7%
  66. 66 Kimi K2 Jul 2025 69.6%
  67. 67 MiMo-7B-RL 68.2%
  68. 68 Llama-3.3-Nemotron-Super-49B-v1 67.5%
  69. 69 DeepSeek-V3 59.4%
  70. 70 Phi-4-mini-reasoning 57.5%
  71. 71 o1-preview 56.7%
  72. 72 ERNIE-4.5-300B-A47B 54.8%
  73. 73 Phi-4-mini-flash-reasoning 52.3%
  74. 74 GPT-4.1 mini 49.6%
  75. 75 Qwen3-1.7B 48.3%
  76. 76 Hunyuan-TurboS 43.3%
  77. 77 GPT-4.5 36.7%
  78. 78 Mercury 30%
  79. 79 GPT-4.1 nano 29.4%
  80. 80 DeepSeek-R1-Distill-Qwen-1.5B 28.9%
  81. 81 Claude 3.5 Sonnet (October 2024) 27.6%
  82. 82 Falcon-H1-34B 23.8%
  83. 83 Cohere Command A 23.3%
  84. 84 InternLM3-8B-Instruct 20%
  85. 85 Phi-4 Mini 10%
  86. 86 Granite 3.3 8B Instruct 8.12%
  87. 87 Claude 3.5 Haiku 5.30%
#ModelScoreSettingDateSource
1 gpt-oss-120b 96.6% Aug 2025 Model card
2 gpt-oss-20b 96% Aug 2025 Model card
3 Grok-3 mini 95.8% Feb 2025 Model card
4 o4-mini 93.4% Apr 2025 Model card
5 Grok 3 93.3% Feb 2025 Model card
6 LongCat-Flash-Thinking 93.3% Sep 2025 Model card
7 DeepSeek-V3.1 93.1% Aug 2025 Model card
8 o3-pro 93% Jun 2025 Model card
9 Gemini 2.5 Pro Mar 2025 92% Mar 2025 Model card
10 Magistral Medium 1.2 91.8% Sep 2025 Model card
11 seed-oss-36b-instruct 91.7% Aug 2025 Model card
12 Solar Open 100B 91.7% Dec 2025 Model card
13 o3 91.6% Apr 2025 Model card
14 DeepSeek-R1 91.4% May 2025 Model card
15 GLM-4.5 91% Jul 2025 Model card
16 STEP3-VL-10B 90.9% Jan 2026 Model card
17 INTELLECT-3 90.8% Nov 2025 Model card
18 Magistral Medium 1.0 90% Jun 2025 Model card
19 Seed1.6-Thinking 90% Jun 2025 Model card
20 Ministral 3 14B 89.8% Dec 2025 Model card
21 GLM-4.5-Air 89.4% Aug 2025 Model card
22 qwen3-coder-next 89% Feb 2026 Model card
23 Falcon H1R 7B 88.1% Jan 2026 Model card
24 Gemini 2.5 Flash Apr 2025 88% Apr 2025 Model card
25 Llama-3.3-Nemotron-Super-49B-v1.5 87.5% Jul 2025 Model card
26 o3-mini 87.3% Jan 2025 Model card
27 Hunyuan-A13B-Instruct 87.3% Jun 2025 Model card
28 Seed-Thinking-v1.5 86.7% Apr 2025 Model card
29 Magistral Small 1.2 86.1% Sep 2025 Model card
30 MiniMax-M1-80k 86% Jun 2025 Model card
31 Qwen3-235B-A22B 85.7% Apr 2025 Model card
32 InternVL3.5-241B-A28B 84.7% Aug 2025 Model card
33 Baichuan-M2-32B 83.4% Aug 2025 Model card
34 nvidia-nemotron-nano-9b-v2 81.9% Aug 2025 Model card
35 Hermes 4 405B 81.9% Aug 2025 Model card
36 Qwen3-32B 81.4% May 2025 Model card
37 Phi-4-Reasoning-plus 81.3% Apr 2025 Model card
38 GLM-Z1-32B-0414 80.8% Apr 2025 Model card
39 Llama-3.1-Nemotron-Ultra-253B-v1 80.8% May 2025 Model card
40 Olmo 3.1 32B Think 80.6% Dec 2025 Model card
41 Qwen3-30B-A3B 80.4% May 2025 Model card
42 Claude 3.7 Sonnet 80% Feb 2025 Model card
43 Qwen3-14B 79.3% May 2025 Model card
44 o1 79.2% Dec 2024 Model card
45 INTELLECT-2 78.8% May 2025 Model card
46 Hunyuan-T1 78.2% Mar 2025 Model card
47 Kimi k1.5 77.5% Jan 2025 Model card
48 Olmo 3 32B Think 76.8% Nov 2025 Model card
49 Solar Pro 2 76.5% Jul 2025 Model card
50 Qwen3-8B 76% May 2025 Model card
51 Code World Model (CWM) 76% Sep 2025 Model card
52 Phi-4-reasoning 75.3% Apr 2025 Model card
53 Qwen3-4B 73.8% May 2025 Model card
54 Gemini 2.0 Flash Thinking Jan 2025 73.3% Jan 2025 Model card
55 LongCat-Flash-Omni 72.9% Oct 2025 Model card
56 DeepSeek-R1-Distill-Qwen-32B 72.6% Jan 2025 Model card
57 EXAONE Deep 32B 72.1% Mar 2025 Model card
58 Magistral Medium 1.1 72% Jul 2025 Model card
59 Hermes 4.3 36B 71.9% Dec 2025 Model card
60 Magistral Small 1.0 70.7% Jun 2025 Model card
61 Magistral Small 1.1 70.5% Jul 2025 Model card
62 LongCat-Flash-Chat 70.4% Aug 2025 Model card
63 DeepSeek-R1-Distill-Llama-70B 70% Jan 2025 Model card
64 Reka Flash 3 70% Mar 2025 Model card
65 DeepSeek-R1-Distill-Qwen-14B 69.7% Jan 2025 Model card
66 Kimi K2 Jul 2025 69.6% Jul 2025 Model card
67 MiMo-7B-RL 68.2% Apr 2025 Model card
68 Llama-3.3-Nemotron-Super-49B-v1 67.5% May 2025 Model card
69 DeepSeek-V3 59.4% Mar 2025 Model card
70 Phi-4-mini-reasoning 57.5% Apr 2025 Model card
71 o1-preview 56.7% Sep 2024 Model card
72 ERNIE-4.5-300B-A47B 54.8% Jun 2025 Model card
73 Phi-4-mini-flash-reasoning 52.3% Jul 2025 Model card
74 GPT-4.1 mini 49.6% Apr 2025 Model card
75 Qwen3-1.7B 48.3% May 2025 Model card
76 Hunyuan-TurboS 43.3% Feb 2025 Model card
77 GPT-4.5 36.7% Feb 2025 Model card
78 Mercury 30% Jun 2025 Model card
79 GPT-4.1 nano 29.4% Apr 2025 Model card
80 DeepSeek-R1-Distill-Qwen-1.5B 28.9% Jan 2025 Model card
81 Claude 3.5 Sonnet (October 2024) 27.6% Oct 2024 Model card
82 Falcon-H1-34B 23.8% May 2025 Model card
83 Cohere Command A 23.3% Apr 2025 Model card
84 InternLM3-8B-Instruct 20% Jan 2025 Model card
85 Phi-4 Mini 10% Mar 2025 Model card
86 Granite 3.3 8B Instruct 8.12% Apr 2025 Model card
87 Claude 3.5 Haiku 5.30% Oct 2024 Model card

6 models have held number one since its release in Feb 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI DeepSeek xAI
Lab claims Passed 85% in Jan 2025, by o3-mini
gpt-oss-120b 96.6%

The best score passed 85% in Jan 2025

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) gpt-oss-120b Grok-3 mini o3-mini DeepSeek-R1 o1-preview 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) gpt-oss-120b o3-mini

Labs headline it less often now, last in Jan 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

OpenAI
Anthropic
DeepSeek
Shanghai AI Lab
Moonshot AI
Google DeepMind
xAI
Tencent
Qwen
Reka
LG AI Research
ByteDance Seed
Z.ai
IBM
Xiaomi Corp
Microsoft
Prime Intellect
Technology Innovation Institute
Mistral AI
MiniMax
Inception Labs
Baidu
Upstage
NVIDIA
Baichuan
Nous Research
Meituan
Meta
Ai2
DateLaunch postScoreAs the post writes it
5 Jan 2026 Falcon H1R 7B 88.1 AIME-24
31 Dec 2025 Solar Open 100B 91.7 AIME 2024
12 Dec 2025 Olmo 3.1 32B Think 80.6 AIME 2024
3 Dec 2025 Hermes 4.3 36B 71.9 AIME 24
2 Dec 2025 Ministral 3 14B 89.8 AIME24 0.898 (reasoning)
26 Nov 2025 INTELLECT-3 90.8 AIME24
20 Nov 2025 Olmo 3 32B Think 76.8 AIME 2024
31 Oct 2025 LongCat-Flash-Omni 72.9 AIME24(avg@10)
24 Sep 2025 Code World Model (CWM) 76 AIME24
21 Sep 2025 LongCat-Flash-Thinking 93.3 AIME24(Mean@32)
17 Sep 2025 Magistral Small 1.2 86.1 AIME24 pass@1
17 Sep 2025 Magistral Medium 1.2 91.8 AIME24 pass@1
29 Aug 2025 LongCat-Flash-Chat 70.4 AIME24(avg@10)
26 Aug 2025 Hermes 4 405B 81.9 AIME'24 (reasoning)
25 Aug 2025 InternVL3.5-241B-A28B 84.7 AIME24
21 Aug 2025 DeepSeek-V3.1 93.1 AIME 2024 (Pass@1)
20 Aug 2025 seed-oss-36b-instruct 91.7 AIME24
10 Aug 2025 Baichuan-M2-32B 83.4 AIME24
5 Aug 2025 gpt-oss-20b 96 AIME 2024 (tools)
5 Aug 2025 gpt-oss-120b 96.6 AIME 2024 (tools)
28 Jul 2025 GLM-4.5 91 AIME24
25 Jul 2025 Llama-3.3-Nemotron-Super-49B-v1.5 87.5 AIME 2024 (pass@1, avg of 16)
24 Jul 2025 Magistral Small 1.1 70.5 AIME24 pass@1
24 Jul 2025 Magistral Medium 1.1 72 AIME24 pass@1
11 Jul 2025 Kimi K2 69.6 AIME 2024, Avg@64
10 Jul 2025 Solar Pro 2 76.5 AIME'24 pass@1
9 Jul 2025 Phi-4-mini-flash-reasoning 52.3 AIME24
30 Jun 2025 ERNIE-4.5-300B-A47B 54.8 AIME'24
27 Jun 2025 Hunyuan-A13B-Instruct 87.3 AIME 2024
26 Jun 2025 Mercury 30 AIME 2024 (% accuracy)
25 Jun 2025 Seed1.6-Thinking 90 AIME2024
16 Jun 2025 MiniMax-M1-80k 86 AIME 2024
10 Jun 2025 Magistral Small 1.0 70.7 AIME24 pass@1
10 Jun 2025 o3-pro 93 Competition Math (AIME 2024), pass@1
10 Jun 2025 Magistral Medium 1.0 90 AIME-24 (maj@64; maj@4 = 81.3, pass@1 = 73.6)
28 May 2025 DeepSeek-R1 91.4 AIME 2024 (Pass@1)
20 May 2025 Falcon-H1-34B 23.8 AIME-24
11 May 2025 INTELLECT-2 78.8 AIME24
30 Apr 2025 Phi-4-reasoning 75.3 AIME 24
30 Apr 2025 Phi-4-Reasoning-plus 81.3 AIME 24
29 Apr 2025 MiMo-7B-RL 68.2 AIME 2024 (Pass@1)
29 Apr 2025 Qwen3-235B-A22B 85.7 AIME’24
17 Apr 2025 Gemini 2.5 Flash 88 AIME 2024 single attempt
16 Apr 2025 o4-mini 93.4 AIME 2024 (no tools)
16 Apr 2025 Granite 3.3 8B Instruct 8.12 AIME24
16 Apr 2025 o3 91.6 AIME 2024 Competition Math, o3 (no tools)
14 Apr 2025 GPT-4.1 nano 29.4 AIME '24
14 Apr 2025 GPT-4.1 mini 49.6 AIME '24
14 Apr 2025 GLM-Z1-32B-0414 80.8 AIME24
10 Apr 2025 Seed-Thinking-v1.5 86.7 AIME 2024
25 Mar 2025 Gemini 2.5 Pro 92 AIME 2024, single attempt (pass@1)
24 Mar 2025 DeepSeek-V3 59.4 AIME
21 Mar 2025 Hunyuan-T1 78.2 AIME 2024
18 Mar 2025 EXAONE Deep 32B 72.1 AIME 2024 pass@1 72.1 (cons@64 90.0)
10 Mar 2025 Reka Flash 3 70 AIME'24 (cons@64)
6 Mar 2025 QwQ-32B AIME24
27 Feb 2025 Hunyuan-TurboS 43.3 AIME2024
27 Feb 2025 GPT-4.5 36.7 AIME '24 (math)
24 Feb 2025 Claude 3.7 Sonnet 80 High school math competition AIME 2024 (64K extended thinking, parallel test-time compute)
19 Feb 2025 Grok-3 mini 95.8 AIME'24 (cons@64; pass@1 = 89.5) [Grok 3 mini Think]
19 Feb 2025 Grok 3 93.3 AIME'24 (cons@64; pass@1 = 83.9)
31 Jan 2025 o3-mini 87.3 Competition Math (AIME 2024)
21 Jan 2025 Gemini 2.0 Flash Thinking 73.3 AIME2024 (Math)
20 Jan 2025 Kimi k1.5 77.5 AIME 2024 (Pass@1), long-CoT
20 Jan 2025 DeepSeek-R1 79.8 AIME 2024 (Pass@1)
13 Jan 2025 InternLM3-8B-Instruct 20 AIME2024(0-shot), Thinking Mode
26 Dec 2024 DeepSeek-V3 39.2 AIME 2024 (Pass@1)
17 Dec 2024 o1 79.2 AIME 2024 (pass @1)
22 Oct 2024 Claude 3.5 Sonnet (October 2024) 27.6 AIME 2024 (Maj@64 0-shot CoT)
12 Sep 2024 o1-preview 56.7 AIME (2024) cons@64

57 system cards report it, from 27 labs

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
28 Feb 2026 qwen3-coder-next technical report 89 AIME24 (Table 9) Capabilities
14 Jan 2026 STEP3-VL-10B technical report 90.9 AIME2024, SeRe (Tables 2 and 3). PaCoRe 93.33 Capabilities
13 Jan 2026 Ministral 3 14B technical report 89.8 AIME 2024, reported as pass@16 (Table 5) Capabilities
11 Jan 2026 Solar Open 100B technical report 91.7 AIME 2024 (Table 6) Capabilities
5 Jan 2026 Falcon H1R 7B technical report 88.1 AIME24 pass@1 (Table 4). DeepConf@512 96.7 (Table 7) Capabilities
18 Dec 2025 INTELLECT-3 technical report 90.8 AIME24, Avg@32 (Table 2, Appendix A.1) Capabilities
15 Dec 2025 Olmo 3 32B Think technical report 76.8 AIME 2024, Avg@32, Final Think 3.0 column (Table 14) Capabilities
15 Dec 2025 Olmo 3.1 32B Think technical report 80.6 AIME 2024, Avg@32 (Tables 1 and 14) Capabilities
31 Oct 2025 LongCat-Flash-Omni technical report 72.9 AIME24 (avg@10) (Table 13) Capabilities
30 Sep 2025 Code World Model (CWM) technical report 76 AIME24, pass@1 averaged over 20 samples (Table 12) Capabilities
23 Sep 2025 LongCat-Flash-Thinking technical report 93.3 AIME-24(Mean@32) (Table 2) Capabilities
2 Sep 2025 Baichuan-M2-32B technical report 83.4 AIME24, max_tokens 64k (Table 1) Capabilities
1 Sep 2025 LongCat-Flash-Chat technical report 70.4 AIME24 (avg@10) (Table 3) Capabilities
25 Aug 2025 InternVL3.5-241B-A28B technical report 84.7 AIME24 (Table 2). OpenCompass run gives 84.1 (Table 14) Capabilities
25 Aug 2025 Hermes 4 405B technical report 81.9 AIME'24, reasoning mode (Table 3). Non-reasoning 11.4 Capabilities
20 Aug 2025 nvidia-nemotron-nano-9b-v2 technical report 81.9 AIME24, NVIDIA-Nemotron-Nano-9B-v2 (Figure 1) Capabilities
8 Aug 2025 GLM-4.5 technical report 91 AIME 24, Avg@32 (Table 4) Capabilities
8 Aug 2025 GLM-4.5-Air technical report 89.4 AIME 24, Avg@32 (Table 4) Capabilities
5 Aug 2025 gpt-oss-20b model card 96 AIME 2024 (with tools), high reasoning (no tools 92.1) (Table 3, Figure 1) Capabilities
5 Aug 2025 gpt-oss-120b model card 96.6 AIME 2024 (with tools), high reasoning (no tools 95.8) (Table 3, Figure 1) Capabilities
30 Jul 2025 Falcon-H1-34B technical report 23.8 AIME-24, average accuracy over 16 repetitions (Table 24) Capabilities
21 Jul 2025 Kimi K2 technical report 69.6 AIME 2024 (Avg@64) (Table 3) Capabilities
9 Jul 2025 Phi-4-mini-flash-reasoning technical report 52.3 AIME24, pass@1 averaged over 64 samples, 32K generation (Table 4) Capabilities
29 Jun 2025 ERNIE-4.5-300B-A47B technical report 54.8 AIME'24 (Table 5) Capabilities
26 Jun 2025 Hunyuan-A13B-Instruct technical report 87.3 AIME2024, slow-thinking (Table 3). Fast-thinking 30.6 (Table 4) Capabilities
16 Jun 2025 MiniMax-M1-80k technical report 86 AIME 2024, avg of 32 samples (Table 2) Capabilities
12 Jun 2025 Magistral Small 1.0 technical report 70.7 AIME'24 pass@1, SFT + RL (Magistral Small) column (Table 3). maj@64 83.3 Capabilities
12 Jun 2025 Magistral Medium 1.0 technical report 73.6 AIME'24, pass@1 averaged over 64 runs (Table 2). maj@64 90.0 Capabilities
14 May 2025 Qwen3-8B technical report 76 AIME’24, thinking mode (Table 17). Non-thinking 29.1 (Table 18) Capabilities
14 May 2025 Qwen3-4B technical report 73.8 AIME’24, thinking mode (Table 17). Non-thinking 25.0 (Table 18) Capabilities
14 May 2025 Qwen3-32B technical report 81.4 AIME’24, thinking mode (Table 13). Non-thinking 31.0 (Table 14) Capabilities
14 May 2025 Qwen3-30B-A3B technical report 80.4 AIME’24, thinking mode (Table 15). Non-thinking 32.8 (Table 16) Capabilities
14 May 2025 Qwen3-235B-A22B technical report 85.7 AIME’24, thinking mode (Table 11). Non-thinking 40.1 (Table 12) Capabilities
14 May 2025 Qwen3-14B technical report 79.3 AIME’24, thinking mode (Table 15). Non-thinking 31.7 (Table 16) Capabilities
14 May 2025 Qwen3-1.7B technical report 48.3 AIME’24, thinking mode (Table 19). Non-thinking 13.4 (Table 20) Capabilities
12 May 2025 MiMo-7B-RL technical report 68.2 AIME 2024 (Pass@1), average of 32 runs (Table 4, Table 5, Table 6) Capabilities
12 May 2025 INTELLECT-2 technical report 78.8 AIME24, evalchemy defaults, target length 10,000 tokens (Table 1) Capabilities
2 May 2025 Llama-3.3-Nemotron-Super-49B-v1 technical report 67.5 AIME24, reasoning on, reasoning off 16.7 (Table 4) Capabilities
2 May 2025 Llama-3.1-Nemotron-Ultra-253B-v1 technical report 80.8 AIME24, reasoning on, reasoning off 20.0 (Table 5) Capabilities
30 Apr 2025 Phi-4-reasoning technical report 74.6 AIME 24, average pass@1, std 5.1 (Table 1) Capabilities
30 Apr 2025 Phi-4-Reasoning-plus technical report 81.3 AIME 24, average pass@1 over 5 runs (Table 1) Capabilities
30 Apr 2025 Phi-4-mini-reasoning technical report 57.5 AIME (AIME24 per Section 5.1), pass@1 averaged over 3 runs (Table 3) Capabilities
29 Apr 2025 Gemini 2.5 Flash model card 88 AIME 2024, single attempt (pass@1) (Table, p.4) Capabilities
28 Apr 2025 Gemini 2.5 Pro model card 92 AIME 2024, single attempt pass@1 (Table, p.4, column Gemini 2.5 Pro Experimental 03-25) Capabilities
10 Apr 2025 Seed-Thinking-v1.5 technical report 86.7 AIME 2024, mean of 32 samples (Table 2) Capabilities
1 Apr 2025 Cohere Command A technical report 23.3 AIME (2024) (Table 15) Capabilities
16 Mar 2025 EXAONE Deep 32B technical report 72.1 AIME 2024, pass@1 with k=64, cons@64 90.0 (Table 2) Capabilities
3 Mar 2025 Phi-4 Mini technical report 10 AIME 2024, Phi-4-Mini row of the CoT reasoning comparison (Table 9) Capabilities
20 Jan 2025 Kimi k1.5 technical report 77.5 AIME 2024 (Pass@1), long-CoT (Table 2). Short-CoT 60.8 (Table 3) Capabilities
20 Jan 2025 DeepSeek-R1 technical report 79.8 AIME 2024 (Pass@1), DeepSeek-R1 (Table 8, same as final column of Table 3) Capabilities
20 Jan 2025 DeepSeek-R1-Distill-Qwen-32B technical report 72.6 AIME 2024 pass@1, cons@64 83.3, DeepSeek-R1-Distill-Qwen-32B (Table 15) Capabilities
20 Jan 2025 DeepSeek-R1-Distill-Qwen-14B technical report 69.7 AIME 2024 pass@1, cons@64 80.0, DeepSeek-R1-Distill-Qwen-14B (Table 15) Capabilities
20 Jan 2025 DeepSeek-R1-Distill-Qwen-1.5B technical report 28.9 AIME 2024 pass@1, cons@64 52.7, DeepSeek-R1-Distill-Qwen-1.5B (Table 15) Capabilities
20 Jan 2025 DeepSeek-R1-Distill-Llama-70B technical report 70 AIME 2024 pass@1, cons@64 86.7, DeepSeek-R1-Distill-Llama-70B (Table 15) Capabilities
26 Dec 2024 DeepSeek-V3 technical report 39.2 AIME 2024 (Pass@1), DeepSeek-V3 (Table 6) Capabilities
22 Oct 2024 Claude 3.5 Sonnet (October 2024) model card 16 AIME 2024, 0-shot CoT. Maj@64 0-shot CoT 27.6% (Table 7) Capabilities
22 Oct 2024 Claude 3.5 Haiku model card 5.3 AIME 2024, 0-shot CoT. Maj@64 0-shot CoT 10.1% (Table 8) Capabilities

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI DeepSeek xAI
AIME 2026 Passed 85% in Feb 2026, by GPT-5.2
GPT-5.5 100%
AIME 2024 Passed 85% in Jan 2025, by o3-mini
gpt-oss-120b 96.6%
AIME 2025 Passed 85% in Feb 2025, by o3-mini
Grok 4 Heavy 100%
AIME 2026 Passed 85% in Feb 2026, by GPT-5.2
GPT-5.5 100%
Replaces AIME 2023 Replaced by AIME 2025

More in Math

All 105 in Math
  • IMO-AnswerBench In use 400 olympiad problems with short verifiable answers, robustified against memorization. Scored by answer match. GLM-5.2 91%
  • HMMT 2026 In use Models solve the 33 final-answer problems from the February 2026 Harvard-MIT Mathematics Tournament. Scored by exact answer match. GPT-5.5 98.5%
  • MathArena (Apex, IMO, USAMO) In use Evaluates models on newly released math competitions after release. Final-answer or human/LLM-judged proof scores. Claude Opus 4.8 81.3%
  • PolyMath In use Models solve 9,000 parallel math problems in 18 languages across four difficulty levels, from school math to frontier math. Final answers are checked. Qwen3.7-Max 86.5%
  • USAMO 2026 In use Models write full proofs for the six problems of the 2026 USA Mathematical Olympiad. Judges grade each proof from 0 to 7 against rubrics. GPT-5.5 98.2%
  • ArXivMath In use Final-answer research-level math problems extracted monthly from recent arXiv papers and filtered to be self-contained. GPT-6.1 Sol 94.2%