MMLU-Pro benchmark

266 AI models ranked on MMLU-Pro. Each row shows the model’s score next to its cheapest API price per 1M tokens where it has one, so you can weigh quality against cost.

Top model today

Gemini 3 Pro (Google) leads at 90%, from $2.00 per 1M input.

Data via arXiv:2406.01574

Models ranked

266

on this benchmark

Top score

90%

current leader

Type

Single eval

benchmark

Updated

15 July 2026

last source fetch

Cost vs MMLU-Pro

160 priced models
Most attractive quadrant (cheap + high)
= reasoning model (extended thinking)

Each labelled name links to that model’s page. Curated to the cost-efficiency frontier, top scorers and cheapest so labels stay readable; the full field is in the leaderboard table below (and via “select all”). Reasoning models (extended “thinking” before answering) carry a ring around the dot.

Models ranked on MMLU-Pro

highest score first · 266 models
Maker
#
1Gemini 3 ProGoogle90%$2.00$12.00
2Claude Opus 4.5Anthropic89%reasoning: true$5.00$25.00
3Gemini 3 FlashGoogle89%reasoning: truen/an/a
4Claude Opus 4.1Anthropic88%reasoning: true$15.00$75.00
5MiniMax M2.1MiniMax87%$0.27$1.20
6Claude 4.5 SonnetAnthropic87%reasoning: true$3.00$15.00
7GPT-5.2OpenAI87%$1.75$14.00
8Claude 4 OpusAnthropic87%reasoning: true$5.00$25.00
9GPT-5OpenAI87%$1.25$10.00
10GPT-5.1OpenAI87%$1.25$10.00
11Grok 4xAI87%$3.00$15.00
12GPT-5 CodexOpenAI86%$1.25$10.00
13DeepSeek V3 2 SpecialeDeepSeek86%$0.58$1.68
14DeepSeek V3.2DeepSeek86%reasoning: true$0.27$0.40
15Gemini 2.5 ProGoogle86%$1.25$10.00
16GPT-5.1 CodexOpenAI86%n/an/a
17Gemini 2.5 Pro (Mar)Google86%n/an/a
18GLM 4.7Zhipu86%$0.40$1.50
19Doubao Seed CodeBytedance-seed85%n/an/a
20Grok 4.1 Fast ReasoningxAI85%$0.20$0.50
21o3OpenAI85%$2.00$8.00
22DeepSeek V3.1 TerminusDeepSeek85%reasoning: true$0.27$1.00
23DeepSeek V3 1DeepSeek85%reasoning: true$0.27$1.00
24Grok 4 Fast ReasoningxAI85%$0.20$0.50
25DeepSeek V3.2 ExpDeepSeek85%n/an/a
26Cogito 671B V2 P1Other85%reasoning: true$1.20$1.20
27R1DeepSeek85%$0.40$0.40
28Kimi K2 ThinkingMoonshot85%$0.60$1.20
29DeepSeek R1 (Jan)DeepSeek84%n/an/a
30Mimo V2 FlashXiaomi84%reasoning: true$0.10$0.30
31Qwen3 235B A22b Instruct 2507Alibaba84%reasoning: true$0.09$0.58
32Gemini 2.5 Flash Preview 09 2025Google84%reasoning: true$0.30$2.50
33Claude 4 SonnetAnthropic84%reasoning: true$3.00$15.00
34Qwen3 MaxAlibaba84%$2.11$8.45
35o1OpenAI84%$15.00$60.00
36Qwen3 Max (Preview)Alibaba84%n/an/a
37K-EXAONELg-ai-research84%n/an/a
38GPT-5 MiniOpenAI84%$0.25$2.00
39Gemini 2.5 Pro (May)Google84%n/an/a
40Claude 3.7 SonnetAnthropic84%reasoning: true$3.00$15.00
41DeepSeek V3.2 ExpDeepSeek84%n/an/a
42Qwen3 VL 235B A22bAlibaba84%reasoning: true$0.53$2.66
43GLM 4.5Zhipu84%$0.40$1.60
44o4 MiniOpenAI83%$1.00$4.00
45Gemini 2.5 FlashGoogle83%reasoning: true$0.15$0.60
46ERNIE 5.0 Thinking PreviewBaidu83%n/an/a
47Nova 2.0 Pro PreviewAmazon83%mediumn/an/a
48GLM 4.6Zhipu83%reasoning: true$0.40$1.75
49Grok 3 MinixAI83%reasoning: true$0.25$0.50
50Llama 3.1 Nemotron Ultra 253BNVIDIA82%reasoning: true$0.60$1.80
150 of 266
Page 1 of 6
Prices are on-demand list rates. Enterprise commitments can be materially lower. Cost per task uses best-effort token usage.

Frequently asked questions

What is the MMLU-Pro benchmark?

A harder, cleaned-up successor to MMLU testing multi-task knowledge and reasoning across academic subjects. It expands each question from four to ten answer options and removes trivial or noisy items, making it more reasoning-intensive and prompt-stable. Scored as multiple-choice accuracy.

Which AI model scores highest on MMLU-Pro?

Gemini 3 Pro (Google) leads with 90%, from $2.00 per 1M input tokens.

How many models are ranked on MMLU-Pro?

266 models carry a MMLU-Pro score. Scores come from Artificial Analysis; where a model has a live API price it is verified against LiteLLM and OpenRouter, and models with no current price are listed without one.

Other benchmarks

compare the same models on a different eval
Benchmark scores via Artificial Analysis · pricing verified against LiteLLM + OpenRouter · refreshed daily · updated 15 July 2026

Every weekday

AI moves fast. Here's your debrief.

News, analysis, tools, and more.

For people who build with AI