MATH / AIME
Competition mathematics, from the OTIS Mock AIME 2024-2025 set. Aggregated from Epoch AI's AI Benchmarking Hub (otis_mock_aime_2024_2025.csv, column 'mean_score'), used under CC-BY 4.0. The Known Good aggregates published results; it runs no evaluations.
142 models scored
Higher is better
Published by Epoch AI
CC-BY 4.0
MATH / AIME
MATH / AIME as measured and published by Epoch AI (CC-BY 4.0). Ingested, normalised only where it feeds an index, and never re-run by us. Higher is better.
16 of 142 models ▾
☰⚙⇩
+Add model from specific provider
💡 Reasoning models are indicated by a lightbulb
The Known Good
Every model we hold a MATH / AIME score for
142 of 142 models
| # | Model | Creator | MATH / AIME | Known Good Index | Measured |
|---|---|---|---|---|---|
| 1 |
OpenAI: GPT-5.6 Sol
💡 |
OpenAI | 100.0 | — | 25 Jul 2026 |
| 2 |
gpt-5.5-pro-pre-release
|
OpenAI | 100.0 | — | 25 Jul 2026 |
| 3 |
gpt-5.5-pre-release
|
OpenAI | 100.0 | 98 | 25 Jul 2026 |
| 4 |
OpenAI: GPT-5.6 Terra
💡 |
OpenAI | 99.7 | — | 25 Jul 2026 |
| 5 |
Anthropic: Claude Fable 5
💡 |
Anthropic | 99.7 | — | 25 Jul 2026 |
| 6 |
Claude Opus 5
💡 |
Anthropic | 98.9 | — | 25 Jul 2026 |
| 7 |
OpenAI: GPT-5.6 Luna
💡 |
OpenAI | 98.3 | — | 25 Jul 2026 |
| 8 |
Anthropic: Claude Opus 4.8
💡 |
Anthropic | 98.3 | — | 25 Jul 2026 |
| 9 |
Anthropic: Claude Opus 4.7
💡 |
Anthropic | 97.8 | 94 | 25 Jul 2026 |
| 10 |
xAI: Grok 4.5
💡 |
xAI | 97.8 | — | 25 Jul 2026 |
| 11 |
MoonshotAI: Kimi K3
💡 |
Moonshot AI | 97.2 | — | 25 Jul 2026 |
| 12 |
DeepSeek: DeepSeek V4 Pro
💡 |
DeepSeek | 96.7 | 93 | 25 Jul 2026 |
| 13 |
MoonshotAI: Kimi K2.7 Code
💡 |
Moonshot AI | 96.4 | — | 25 Jul 2026 |
| 14 |
MoonshotAI: Kimi K2.6
💡 |
Moonshot AI | 96.1 | 93 | 25 Jul 2026 |
| 15 |
OpenAI: GPT-5.2
💡 |
OpenAI | 96.1 | 80 | 25 Jul 2026 |
| 16 |
Google: Gemini 3.1 Pro Preview
💡 |
95.6 | 96 | 25 Jul 2026 | |
| 17 |
Google: Gemini 3.5 Flash
💡 |
95.6 | 95 | 25 Jul 2026 | |
| 18 |
OpenAI: GPT-5.4
💡 |
OpenAI | 95.3 | 90 | 25 Jul 2026 |
| 19 |
Qwen: Qwen3.7 Max
💡 |
Alibaba | 95.0 | 93 | 25 Jul 2026 |
| 20 |
Anthropic: Claude Sonnet 5
💡 |
Anthropic | 94.7 | — | 25 Jul 2026 |
| 21 |
claude-opus-4-6_64K
|
Anthropic | 94.4 | — | 25 Jul 2026 |
| 22 |
xAI: Grok 4.3
💡 |
xAI | 93.3 | — | 25 Jul 2026 |
| 23 |
Anthropic: Claude Opus 4.6
💡 |
Anthropic | 93.1 | 88 | 25 Jul 2026 |
| 24 |
Google: Gemini 3 Flash Preview
💡 |
92.8 | 83 | 25 Jul 2026 | |
| 25 |
Z.ai: GLM 5.1
💡 |
Z.ai | 92.2 | 88 | 25 Jul 2026 |
| 26 |
grok-4.20-0309-reasoning
|
xAI | 92.2 | — | 25 Jul 2026 |
| 27 |
fireworks/kimi-k2p5
|
Moonshot AI | 92.2 | — | 25 Jul 2026 |
| 28 |
OpenAI: GPT-5
💡 |
OpenAI | 91.4 | 73 | 25 Jul 2026 |
| 29 |
gemini-3-pro-preview
|
91.4 | 83 | 25 Jul 2026 | |
| 30 |
Qwen: Qwen3.6 Max Preview
💡 |
Alibaba | 91.1 | 90 | 25 Jul 2026 |
| 31 |
Qwen: Qwen3.6 Plus
💡 |
Alibaba | 90.6 | 78 | 25 Jul 2026 |
| 32 |
muse-spark
|
Meta | 88.9 | — | 25 Jul 2026 |
| 33 |
OpenAI: gpt-oss-120b
💡 |
OpenAI | 88.9 | 61 | 25 Jul 2026 |
| 34 |
OpenAI: GPT-5.1
💡 |
OpenAI | 88.6 | 74 | 25 Jul 2026 |
| 35 |
deepseek-reasoner
|
DeepSeek | 87.8 | — | 25 Jul 2026 |
| 36 |
OpenAI: GPT-5.4 Nano
💡 |
OpenAI | 87.8 | — | 25 Jul 2026 |
| 37 |
OpenAI: GPT-5.4 Mini
💡 |
OpenAI | 87.2 | — | 25 Jul 2026 |
| 38 |
OpenAI: GPT-5 Mini
💡 |
OpenAI | 86.7 | 60 | 25 Jul 2026 |
| 39 |
Qwen: Qwen3 235B A22B Thinking 2507
💡 |
Alibaba | 86.7 | — | 25 Jul 2026 |
| 40 |
Z.ai: GLM 5.2
💡 |
Z.ai | 86.4 | 91 | 25 Jul 2026 |
| 41 |
Qwen: Qwen3.6 Flash
💡 |
Alibaba | 86.1 | — | 25 Jul 2026 |
| 42 |
claude-opus-4-5-20251101
|
Anthropic | 86.1 | 72 | 25 Jul 2026 |
| 43 |
Anthropic: Claude Sonnet 4.6
💡 |
Anthropic | 85.8 | 80 | 25 Jul 2026 |
| 44 |
Qwen: Qwen3.5-Flash
💡 |
Alibaba | 85.6 | — | 25 Jul 2026 |
| 45 |
qwen3.5-plus
|
Alibaba | 85.0 | — | 25 Jul 2026 |
| 46 |
Google: Gemini 2.5 Pro
💡 |
84.2 | 64 | 25 Jul 2026 | |
| 47 |
grok-4-0709
|
xAI | 84.0 | 67 | 25 Jul 2026 |
| 48 |
OpenAI: o3
💡 |
OpenAI | 83.9 | 67 | 25 Jul 2026 |
| 49 |
Z.ai: GLM 4.7
💡 |
Z.ai | 83.3 | 68 | 25 Jul 2026 |
| 50 |
kimi-k2-thinking-turbo
|
Moonshot AI | 83.1 | — | 25 Jul 2026 |
| 51 |
OpenAI: o4 Mini
💡 |
OpenAI | 81.7 | — | 25 Jul 2026 |
| 52 |
claude-opus-4-5-20251101_16K
|
Anthropic | 81.7 | — | 25 Jul 2026 |
| 53 |
OpenAI: GPT-5 Nano
💡 |
OpenAI | 81.1 | 57 | 25 Jul 2026 |
| 54 |
Z.ai: GLM 5
💡 |
Z.ai | 80.0 | 77 | 25 Jul 2026 |
| 55 |
claude-sonnet-4-5-20250929_59K
|
Anthropic | 77.8 | — | 25 Jul 2026 |
| 56 |
claude-sonnet-4-5-20250929
|
Anthropic | 77.8 | 59 | 25 Jul 2026 |
| 57 |
grok-3-mini-beta
|
xAI | 77.8 | — | 25 Jul 2026 |
| 58 |
OpenAI: o3 Mini
💡 |
OpenAI | 76.9 | 82 | 25 Jul 2026 |
| 59 |
Qwen: Qwen3 Max
💡 |
Alibaba | 73.3 | — | 25 Jul 2026 |
| 60 |
OpenAI: o1
💡 |
OpenAI | 73.3 | 63 | 25 Jul 2026 |
| 61 |
gemini-2.5-flash-preview-04-17
|
73.1 | — | 25 Jul 2026 | |
| 62 |
claude-sonnet-4-5-20250929_16K
|
Anthropic | 71.1 | — | 25 Jul 2026 |
| 63 |
claude-sonnet-4-20250514
|
Anthropic | 71.1 | — | 25 Jul 2026 |
| 64 |
gemini-2.5-flash-preview-05-20
|
70.8 | — | 25 Jul 2026 | |
| 65 |
claude-opus-4-1-20250805_27K
|
Anthropic | 68.9 | — | 25 Jul 2026 |
| 66 |
claude-sonnet-4-20250514_59K
|
Anthropic | 68.9 | — | 25 Jul 2026 |
| 67 |
claude-haiku-4-5-20251001
|
Anthropic | 66.7 | 56 | 25 Jul 2026 |
| 68 |
DeepSeek: R1 0528
💡 |
DeepSeek | 66.4 | — | 25 Jul 2026 |
| 69 |
deepseek-r1-0528
|
DeepSeek | 66.4 | — | 25 Jul 2026 |
| 70 |
claude-opus-4-1-20250805_16K
|
Anthropic | 64.4 | — | 25 Jul 2026 |
| 71 |
claude-opus-4-20250514_27K
|
Anthropic | 64.4 | — | 25 Jul 2026 |
| 72 |
claude-opus-4-20250514_16K
|
Anthropic | 60.0 | — | 25 Jul 2026 |
| 73 |
claude-3-7-sonnet-20250219_64K
|
Anthropic | 57.8 | — | 25 Jul 2026 |
| 74 | GD gemini-2.0-flash-thinking-exp-01-21 | Google DeepMind,Google | 57.8 | 62 | 25 Jul 2026 |
| 75 |
grok-3-beta
|
xAI | 55.6 | — | 25 Jul 2026 |
| 76 |
DeepSeek: R1
💡 |
DeepSeek | 53.3 | 68 | 25 Jul 2026 |
| 77 |
claude-sonnet-4-20250514_16K
|
Anthropic | 53.3 | — | 25 Jul 2026 |
| 78 |
claude-3-7-sonnet-20250219
|
Anthropic | 53.3 | 70 | 25 Jul 2026 |
| 79 |
deepseek-r1
|
DeepSeek | 53.3 | 68 | 25 Jul 2026 |
| 80 |
DeepSeek: R1 Distill Llama 70B
💡 |
DeepSeek | 51.4 | 52 | 25 Jul 2026 |
| 81 |
o1-mini
|
OpenAI | 46.9 | 56 | 25 Jul 2026 |
| 82 |
claude-3-7-sonnet-20250219_16K
|
Anthropic | 46.7 | — | 25 Jul 2026 |
| 83 |
OpenAI: GPT-4.1 Mini
|
OpenAI | 44.7 | — | 25 Jul 2026 |
| 84 |
claude-opus-4-20250514
|
Anthropic | 42.2 | 62 | 25 Jul 2026 |
| 85 |
claude-opus-4-1-20250805
|
Anthropic | 40.0 | 49 | 25 Jul 2026 |
| 86 |
OpenAI: GPT-4.1
|
OpenAI | 38.3 | 36 | 25 Jul 2026 |
| 87 |
DeepSeek: DeepSeek V3 0324
|
DeepSeek | 37.8 | 60 | 25 Jul 2026 |
| 88 |
gpt-4.5-preview
|
OpenAI | 37.8 | 47 | 25 Jul 2026 |
| 89 |
mistral-medium-2505
|
Mistral | 32.2 | — | 25 Jul 2026 |
| 90 | GD gemini-2.0-flash-001 | Google DeepMind,Google | 31.1 | 53 | 25 Jul 2026 |
| 91 |
o1-preview
|
OpenAI | 31.1 | — | 25 Jul 2026 |
| 92 |
magistral-small-2506
|
Mistral | 30.0 | — | 25 Jul 2026 |
| 93 |
OpenAI: GPT-4.1 Nano
|
OpenAI | 28.9 | — | 25 Jul 2026 |
| 94 |
gemini-1.5-pro-002
|
23.1 | — | 25 Jul 2026 | |
| 95 |
Llama-4-Maverick-17B-128E-Instruct-FP8
|
Meta | 20.6 | — | 25 Jul 2026 |
| 96 |
Google: Gemma 3 27B
|
19.7 | 37 | 25 Jul 2026 | |
| 97 |
gemma-3-27b-it
|
19.7 | 37 | 25 Jul 2026 | |
| 98 |
Qwen: Qwen-Plus
💡 |
Alibaba | 17.8 | — | 25 Jul 2026 |
| 99 |
gemini-1.5-flash-002
|
16.2 | — | 25 Jul 2026 | |
| 100 |
qwen-max
|
Alibaba | 16.1 | — | 25 Jul 2026 |
| 101 |
DeepSeek: DeepSeek V3
|
DeepSeek | 15.8 | 44 | 25 Jul 2026 |
| 102 |
Microsoft: Phi 4
|
Microsoft | 13.8 | 33 | 25 Jul 2026 |
| 103 |
grok-2-1212
|
xAI | 11.5 | 38 | 25 Jul 2026 |
| 104 |
Llama-3.1-405B-Instruct
|
Meta | 9.7 | — | 25 Jul 2026 |
| 105 |
Mistral Large 2407
|
Mistral | 8.5 | — | 25 Jul 2026 |
| 106 |
claude-3-5-sonnet-20241022
|
Anthropic | 8.5 | 40 | 25 Jul 2026 |
| 107 |
Qwen2.5 72B Instruct
|
Alibaba | 8.1 | — | 25 Jul 2026 |
| 108 |
Llama-4-Scout-17B-16E-Instruct
|
Meta | 7.8 | — | 25 Jul 2026 |
| 109 |
mistral-large-2411
|
Mistral | 7.8 | 33 | 25 Jul 2026 |
| 110 |
qwen2.5-32b-instruct
|
Alibaba | 7.4 | — | 25 Jul 2026 |
| 111 |
OpenAI: GPT-4o-mini
|
OpenAI | 6.9 | 23 | 25 Jul 2026 |
| 112 |
gemini-1.5-pro-001
|
6.8 | — | 25 Jul 2026 | |
| 113 |
OpenAI: GPT-4 Turbo
|
OpenAI | 6.7 | — | 25 Jul 2026 |
| 114 |
claude-3-5-sonnet-20240620
|
Anthropic | 6.5 | — | 25 Jul 2026 |
| 115 |
OpenAI: GPT-4o
|
OpenAI | 6.4 | 21 | 25 Jul 2026 |
| 116 |
qwen-turbo
|
Alibaba | 6.1 | — | 25 Jul 2026 |
| 117 |
mistral-small-2503
|
Mistral | 5.8 | 28 | 25 Jul 2026 |
| 118 |
mistral-small-2501
|
Mistral | 5.3 | 26 | 25 Jul 2026 |
| 119 |
Meta: Llama 3.3 70B Instruct
|
Meta | 5.1 | 31 | 25 Jul 2026 |
| 120 |
claude-3-opus-20240229
|
Anthropic | 4.7 | 30 | 25 Jul 2026 |
| 121 |
gemini-1.5-flash-8b-001
|
4.6 | — | 25 Jul 2026 | |
| 122 | AI Llama-3.1-Tulu-3-70B-DPO | Allen Institute for AI,University of Washington | 4.4 | — | 25 Jul 2026 |
| 123 |
claude-3-5-haiku-20241022
|
Anthropic | 4.3 | 23 | 25 Jul 2026 |
| 124 |
Meta-Llama-3-70B-Instruct
|
Meta | 4.3 | — | 25 Jul 2026 |
| 125 |
gemini-1.5-flash-001
|
3.9 | — | 25 Jul 2026 | |
| 126 |
Meta: Llama 3.1 70B Instruct
|
Meta | 3.6 | — | 25 Jul 2026 |
| 127 |
Llama-3.2-90B-Vision-Instruct
|
Meta | 2.6 | — | 25 Jul 2026 |
| 128 |
Meta: Llama 3.1 8B Instruct
|
Meta | 2.5 | — | 25 Jul 2026 |
| 129 |
claude-2.0
|
Anthropic | 2.5 | — | 25 Jul 2026 |
| 130 |
claude-3-sonnet-20240229
|
Anthropic | 2.5 | — | 25 Jul 2026 |
| 131 | NR Hermes-2-Theta-Llama-3-70B | Nous Research,Arcee AI | 2.5 | — | 25 Jul 2026 |
| 132 |
claude-2.1
|
Anthropic | 1.9 | — | 25 Jul 2026 |
| 133 |
mistral-large-2402
|
Mistral | 1.9 | — | 25 Jul 2026 |
| 134 |
claude-3-haiku-20240307
|
Anthropic | 1.8 | — | 25 Jul 2026 |
| 135 |
Google: Gemma 2 27B
|
1.4 | 19 | 25 Jul 2026 | |
| 136 |
gemma-2-27b-it
|
1.4 | 19 | 25 Jul 2026 | |
| 137 |
gpt-4-0613
|
OpenAI | 1.1 | — | 25 Jul 2026 |
| 138 |
gemini-1.0-pro-001
|
1.1 | — | 25 Jul 2026 | |
| 139 |
Meta-Llama-3-8B-Instruct
|
Meta | 0.8 | — | 25 Jul 2026 |
| 140 |
gpt-4-0314
|
OpenAI | 0.6 | — | 25 Jul 2026 |
| 141 |
gemma-2-9b-it
|
0.6 | 10 | 25 Jul 2026 | |
| 142 |
Llama-2-70b-chat-hf
|
Meta | 0.0 | — | 25 Jul 2026 |
The Known Good
Provenance. These figures are published by Epoch AI and redistributed here under CC-BY 4.0.
The publisher's own page is here. We did not run this evaluation and we do not adjust the published numbers — where a score feeds an index it is
min-max normalised against every other model holding the same evaluation, and nothing else is done to it.
This evaluation is one of the six components of the
Known Good Index. Full licence detail is on attribution, and every score here is
in the CSV download.