AIME 2025 — AI model leaderboard

AI models ranked by AIME 2025, an aggregated third-party benchmark from artificial_analysis. Higher is better. Cross-referenced against our first-party meo scores and Effective Value (𝕍).

Ranking 32 models across the full field · as of 2026-06-07.

#	Model	Lab	AIME 2025
1	OpenAI: GPT-5.2 Chat	openai	99.0%
2	OpenAI: GPT-5 Codex	openai	98.7%
3	OpenAI: GPT-5.1-Codex	openai	95.7%
4	OpenAI: GPT-5	openai	94.3%
5	OpenAI: GPT-5.1	openai	94.0%
6	OpenAI: gpt-oss-120b	openai	93.4%
7	OpenAI: GPT-5.1-Codex-Mini	openai	91.7%
8	OpenAI: GPT-5 Mini	openai	90.7%
9	OpenAI: o4 Mini	openai	90.7%
10	OpenAI: gpt-oss-20b	openai	89.3%
11	OpenAI: o3	openai	88.3%
12	Prime Intellect: INTELLECT-3	prime-intellect	88.0%
13	Google: Gemini 2.5 Pro	google	87.7%
14	OpenAI: GPT-5 Nano	openai	83.7%
15	Xiaomi: MiMo-V2-Flash	xiaomi	67.7%
16	Google: Gemini 2.5 Flash	google	60.3%
17	Google: Gemini 3 Flash Preview	google	55.7%
18	Google: Gemini 2.5 Flash Lite Preview 09-2025	google	46.7%
19	OpenAI: GPT-4.1 Mini	openai	46.3%
20	OpenAI: GPT-4.1	openai	34.7%
21	Reka Flash 3	rekaai	33.7%
22	OpenAI: GPT-4.1 Nano	openai	24.0%
23	Google: Gemma 3 27B	google	20.7%
24	Meta: Llama 4 Maverick	meta-llama	19.3%
25	Google: Gemma 3 12B	google	18.3%
26	Microsoft: Phi 4	microsoft	18.0%
27	OpenAI: GPT-4o-mini	openai	14.7%
28	Meta: Llama 4 Scout	meta-llama	14.0%
29	Cohere: Command A	cohere	13.0%
30	Google: Gemma 3 4B	google	12.7%
31	Microsoft: Phi 4 Mini Instruct	microsoft	6.7%
32	OpenAI: GPT-4o	openai	6.0%

Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.

← All rankings Methodology & 𝕍 →