MATH Level 5
Hardest tier of the MATH dataset. Aggregated from Epoch AI's AI Benchmarking Hub (math_level_5.csv, column 'mean_score'), used under CC-BY 4.0. The Known Good aggregates published results; it runs no evaluations.
97 models scored
Higher is better
Published by Epoch AI
CC-BY 4.0
MATH Level 5
MATH Level 5 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.
💡 A lightbulb marks a model its provider publishes as a reasoning model
The Known Good
Every model we hold a MATH Level 5 score for
All 97 of 97 models
| # | Model | Creator | MATH Level 5 | Known Good Index | Measured |
|---|---|---|---|---|---|
| 1 |
OpenAI: GPT-5
💡 |
OpenAI | 98.1 | 78 | 4 Sep 2026 |
| 2 |
OpenAI: GPT-5 Mini
💡 |
OpenAI | 97.8 | 67 | 4 Sep 2026 |
| 3 |
OpenAI: o4 Mini
💡 |
OpenAI | 97.8 | — | 4 Sep 2026 |
| 4 |
OpenAI: o3
💡 |
OpenAI | 97.8 | 74 | 4 Sep 2026 |
| 5 |
claude-sonnet-4-5-20250929
|
Anthropic | 97.7 | 66 | 4 Sep 2026 |
| 6 |
Qwen: Qwen3 Max
💡 |
Alibaba | 97.1 | — | 4 Sep 2026 |
| 7 |
DeepSeek: R1 0528
💡 |
DeepSeek | 96.6 | — | 4 Sep 2026 |
| 8 |
OpenAI: o3 Mini
💡 |
OpenAI | 96.5 | 86 | 4 Sep 2026 |
| 9 |
claude-haiku-4-5-20251001
|
Anthropic | 96.4 | 67 | 4 Sep 2026 |
| 10 |
Google: Gemini 2.5 Pro Preview 05-06
💡 |
95.9 | — | 4 Sep 2026 | |
| 11 |
gemini-2.5-pro-preview-03-25
|
95.6 | — | 4 Sep 2026 | |
| 12 |
OpenAI: GPT-5 Nano
💡 |
OpenAI | 95.2 | 68 | 4 Sep 2026 |
| 13 |
OpenAI: o1
💡 |
OpenAI | 94.7 | 70 | 4 Sep 2026 |
| 14 |
DeepSeek: R1
💡 |
DeepSeek | 93.1 | 75 | 4 Sep 2026 |
| 15 |
claude-3-7-sonnet-20250219_64K
|
Anthropic | 91.2 | — | 4 Sep 2026 |
| 16 |
grok-3-mini-beta
|
xAI | 90.9 | — | 4 Sep 2026 |
| 17 |
claude-3-7-sonnet-20250219
|
Anthropic | 90.0 | 74 | 4 Sep 2026 |
| 18 |
DeepSeek: R1 Distill Llama 70B
💡 |
DeepSeek | 89.9 | 63 | 4 Sep 2026 |
| 19 |
o1-mini
|
OpenAI | 89.2 | 65 | 4 Sep 2026 |
| 20 |
grok-3-beta
|
xAI | 88.7 | — | 4 Sep 2026 |
| 21 |
OpenAI: GPT-4.1 Mini
|
OpenAI | 87.3 | — | 4 Sep 2026 |
| 22 |
DeepSeek-R1-Distill-Qwen-14B
|
DeepSeek | 87.1 | — | 4 Sep 2026 |
| 23 |
claude-3-7-sonnet-20250219_16K
|
Anthropic | 86.3 | — | 4 Sep 2026 |
| 24 |
claude-opus-4-20250514
|
Anthropic | 85.0 | 68 | 4 Sep 2026 |
| 25 |
claude-sonnet-4-20250514
|
Anthropic | 84.4 | — | 4 Sep 2026 |
| 26 |
gemini-2.0-pro-exp-02-05
|
83.5 | 74 | 4 Sep 2026 | |
| 27 |
OpenAI: GPT-4.1
|
OpenAI | 83.0 | 46 | 4 Sep 2026 |
| 28 | GD gemini-2.0-flash-001 | Google DeepMind,Google | 82.2 | 61 | 4 Sep 2026 |
| 29 |
o1-preview
|
OpenAI | 81.6 | — | 4 Sep 2026 |
| 30 |
Mistral: Mistral Medium 3
|
Mistral | 81.6 | — | 4 Sep 2026 |
| 31 |
gpt-4.5-preview
|
OpenAI | 78.6 | 54 | 4 Sep 2026 |
| 32 |
DeepSeek: DeepSeek V3 0324
|
DeepSeek | 75.5 | 64 | 4 Sep 2026 |
| 33 |
Google: Gemma 3 27B
💡 |
74.0 | 47 | 4 Sep 2026 | |
| 34 |
Llama-4-Maverick-17B-128E-Instruct-FP8
|
Meta | 73.0 | — | 4 Sep 2026 |
| 35 |
gemini-1.5-pro-002
|
70.4 | — | 4 Sep 2026 | |
| 36 |
OpenAI: GPT-4.1 Nano
|
OpenAI | 70.0 | — | 4 Sep 2026 |
| 37 |
Qwen: Qwen3 235B A22B
💡 |
Alibaba | 68.9 | — | 4 Sep 2026 |
| 38 |
qwen-max
|
Alibaba | 67.2 | — | 4 Sep 2026 |
| 39 |
Qwen: Qwen-Plus
💡 |
Alibaba | 65.3 | — | 4 Sep 2026 |
| 40 |
Microsoft: Phi 4
|
Microsoft | 64.9 | 41 | 4 Sep 2026 |
| 41 |
DeepSeek: DeepSeek V3
|
DeepSeek | 64.9 | 62 | 4 Sep 2026 |
| 42 |
grok-2-1212
|
xAI | 63.5 | 45 | 4 Sep 2026 |
| 43 |
Qwen2.5 72B Instruct
|
Alibaba | 63.2 | — | 4 Sep 2026 |
| 44 |
Llama-4-Scout-17B-16E-Instruct
|
Meta | 62.3 | — | 4 Sep 2026 |
| 45 |
gemini-1.5-flash-002
|
61.9 | — | 4 Sep 2026 | |
| 46 |
claude-3-5-sonnet-20241022
|
Anthropic | 56.9 | 45 | 4 Sep 2026 |
| 47 |
qwen-turbo
|
Alibaba | 56.2 | — | 4 Sep 2026 |
| 48 |
qwen2.5-32b-instruct
|
Alibaba | 56.1 | — | 4 Sep 2026 |
| 49 |
OpenAI: GPT-4o
|
OpenAI | 53.3 | 27 | 4 Sep 2026 |
| 50 |
OpenAI: GPT-4o-mini
|
OpenAI | 52.6 | 31 | 4 Sep 2026 |
| 51 |
claude-3-5-sonnet-20240620
|
Anthropic | 51.7 | — | 4 Sep 2026 |
| 52 |
mistral-large-2411
|
Mistral | 50.3 | 37 | 4 Sep 2026 |
| 53 |
Llama-3.1-405B-Instruct
|
Meta | 49.8 | — | 4 Sep 2026 |
| 54 |
mistral-small-2503
|
Mistral | 46.8 | 33 | 4 Sep 2026 |
| 55 |
OpenAI: GPT-4 Turbo
|
OpenAI | 46.7 | — | 4 Sep 2026 |
| 56 |
claude-3-5-haiku-20241022
|
Anthropic | 46.4 | 30 | 4 Sep 2026 |
| 57 |
Mistral Large 2407
|
Mistral | 44.8 | — | 4 Sep 2026 |
| 58 |
mistral-small-2501
|
Mistral | 44.8 | 31 | 4 Sep 2026 |
| 59 | AI Llama-3.1-Tulu-3-70B-DPO | Allen Institute for AI,University of Washington | 42.7 | — | 4 Sep 2026 |
| 60 |
Meta: Llama 3.3 70B Instruct
|
Meta | 41.6 | 34 | 4 Sep 2026 |
| 61 |
gemini-1.5-pro-001
|
40.7 | — | 4 Sep 2026 | |
| 62 |
gpt-4-1106-preview
|
OpenAI | 40.0 | — | 4 Sep 2026 |
| 63 |
Llama-3.2-90B-Vision-Instruct
|
Meta | 39.4 | — | 4 Sep 2026 |
| 64 |
qwen2-72b-instruct
|
Alibaba | 39.1 | — | 4 Sep 2026 |
| 65 |
claude-3-opus-20240229
|
Anthropic | 37.5 | 32 | 4 Sep 2026 |
| 66 |
Meta: Llama 3.1 70B Instruct
|
Meta | 36.7 | — | 4 Sep 2026 |
| 67 |
gpt-4-0125-preview
|
OpenAI | 35.4 | — | 4 Sep 2026 |
| 68 |
Google: Gemma 2 27B
|
27.9 | 21 | 4 Sep 2026 | |
| 69 |
WizardLM-2 8x22B
|
Microsoft | 25.7 | — | 4 Sep 2026 |
| 70 | 0A Yi-1.5-34B-Chat | 01.AI | 25.5 | — | 4 Sep 2026 |
| 71 |
gemini-1.5-flash-001
|
25.1 | — | 4 Sep 2026 | |
| 72 |
mistral-large-2402
|
Mistral | 24.5 | — | 4 Sep 2026 |
| 73 |
open-mixtral-8x22b
|
Mistral | 24.2 | — | 4 Sep 2026 |
| 74 |
gpt-4-0613
|
OpenAI | 23.0 | — | 4 Sep 2026 |
| 75 |
Meta: Llama 3.1 8B Instruct
|
Meta | 22.9 | — | 4 Sep 2026 |
| 76 | NR Hermes-2-Theta-Llama-3-70B | Nous Research,Arcee AI | 22.7 | — | 4 Sep 2026 |
| 77 |
Meta-Llama-3-70B-Instruct
|
Meta | 22.6 | — | 4 Sep 2026 |
| 78 |
gemma-2-9b-it
|
21.0 | 13 | 4 Sep 2026 | |
| 79 |
claude-3-sonnet-20240229
|
Anthropic | 18.2 | — | 4 Sep 2026 |
| 80 |
Phi-3-medium-128k-instruct
|
Microsoft | 17.6 | — | 4 Sep 2026 |
| 81 |
gpt-3.5-turbo-1106
|
OpenAI | 15.9 | — | 4 Sep 2026 |
| 82 |
ministral-8b-2410
|
Mistral | 14.9 | — | 4 Sep 2026 |
| 83 |
Anthropic: Claude 3 Haiku
|
Anthropic | 14.9 | — | 4 Sep 2026 |
| 84 |
ministral-3b-2410
|
Mistral | 14.4 | — | 4 Sep 2026 |
| 85 |
claude-2.0
|
Anthropic | 11.7 | — | 4 Sep 2026 |
| 86 | Da dbrx-instruct | Databricks | 11.7 | — | 4 Sep 2026 |
| 87 |
gpt-3.5-turbo-0125
|
OpenAI | 11.6 | — | 4 Sep 2026 |
| 88 |
gemini-1.0-pro-001
|
11.2 | — | 4 Sep 2026 | |
| 89 |
open-mistral-nemo-2407
|
Mistral | 10.8 | — | 4 Sep 2026 |
| 90 |
open-mixtral-8x7b
|
Mistral | 10.0 | — | 4 Sep 2026 |
| 91 |
Mixtral-8x7B-Instruct-v0.1
|
Mistral | 9.3 | — | 4 Sep 2026 |
| 92 |
deepseek-llm-67b-chat
|
DeepSeek | 6.4 | — | 4 Sep 2026 |
| 93 |
Meta-Llama-3-8B-Instruct
|
Meta | 6.1 | — | 4 Sep 2026 |
| 94 | 0A Yi-34B-Chat | 01.AI | 5.1 | — | 4 Sep 2026 |
| 95 |
open-mistral-7b
|
Mistral | 3.7 | — | 4 Sep 2026 |
| 96 |
Mistral-7B-Instruct-v0.3
|
Mistral | 3.6 | — | 4 Sep 2026 |
| 97 |
Llama-2-70b-chat-hf
|
Meta | 3.3 | — | 4 Sep 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 seven components of the
Known Good Index. Full licence detail is on attribution, and every score here is
in the CSV download.