Skip to content
IFAI

Model rankings

505 models · 3 boards · 2 sources

Artificial AnalysisIntelligence IndexArenaText arena

Intelligence Index · A composite of several evaluations · Updated August 18, 2025

Smartest
GPT-5 (high)OpenAI69.0Intelligence
Best open model
Qwen3 235B A22B 2507 (Reasoning)Alibaba63.6Intelligence
Best value
Gemma 3n E4B InstructGoogle739 pts/$Value
Best from China
Qwen3 235B A22B 2507 (Reasoning)Alibaba63.6Intelligence
Top three by Intelligence
1GPT-5 (high)OpenAI69.0
2Grok 4xAI67.5
3o3-proOpenAI67.5
Top three open models
1Qwen3 235B A22B 2507 (Reasoning)Alibaba · #8 overall63.6
2gpt-oss-120BOpenAI · #9 overall61.3
3DeepSeek R1 0528 (May '25)DeepSeek · #13 overall58.7
This week's changes
Mistral Medium 3.1Intelligence · #58New
Gemma 3 4B InstructIntelligence · #132Up 13
Gemini 2.5 Flash-Lite (Reasoning)Intelligence · #32Up 7
GPT-4.1 miniIntelligence · #38Up 7
GPT-4.1 nanoIntelligence · #76Up 3
RankChangeModelIntelligencePrice $/MSpeed tok/sFirst token
1—Qwen3 235B A22B 2507 (Reasoning)OpenAlibaba63.6$2.6——
2—gpt-oss-120BOpenOpenAI61.3$0.26——
3—DeepSeek R1 0528 (May '25)OpenDeepSeek58.7$0.96——
4—GLM-4.5OpenZ AI56.1$0.96——
5—Qwen3 30B A3B 2507 (Reasoning)OpenAlibaba53.2$0.75——
6—MiniMax M1 80kOpenMiniMax53.2$0.83——
7—Llama Nemotron Super 49B v1.5 (Reasoning)OpenNVIDIA52.4$0.000——
8—MiniMax M1 40kOpenMiniMax51.1$0.83——
9—Qwen3 235B A22B 2507 (Non-reasoning)OpenAlibaba50.7$1.2——
10—EXAONE 4.0 32B (Reasoning)OpenLG AI Research50.7$0.70——
11—DeepSeek R1 (Jan '25)OpenDeepSeek49.9$1.8——
12—GLM-4.5-AirOpenZ AI49.5$0.43——
13—gpt-oss-20BOpenOpenAI49.0$0.088——
14—Kimi K2OpenMoonshot AI48.6$1.1——
15—Qwen3 235B A22B (Reasoning)OpenAlibaba48.4$2.6——
16—QwQ 32BOpenAlibaba47.7$0.47——
17—Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)OpenNVIDIA46.4$0.90——
18—Qwen3 30B A3B 2507 (Non-reasoning)OpenAlibaba46.1$0.35——
19▲ 1Qwen3 Coder 480B A35BOpenAlibaba45.2$3.0——
20▲ 1Qwen3 32B (Reasoning)OpenAlibaba44.3$2.6——
21▲ 1DeepSeek V3 0324 (Mar '25)OpenDeepSeek44.0$0.48——
22▲ 1Qwen3 30B A3B (Reasoning)OpenAlibaba42.3$0.75——
23▲ 1Llama 4 MaverickOpenMeta41.7$0.39——
24▲ 1DeepSeek R1 0528 Qwen3 8BOpenDeepSeek41.5$0.065——
25▲ 1DeepSeek R1 Distill Qwen 32BOpenDeepSeek41.2$0.20——
26▲ 1Qwen3 8B (Reasoning)OpenAlibaba40.8$0.66——
27▲ 1Llama 3.3 Nemotron Super 49B v1 (Reasoning)OpenNVIDIA40.5$0.000——
28▼ 9Qwen3 14B (Reasoning)OpenAlibaba40.3$1.3——
29▲ 1DeepSeek R1 Distill Qwen 14BOpenDeepSeek38.2$0.88——
30▲ 1DeepSeek R1 Distill Llama 70BOpenDeepSeek37.4$0.77——
31▼ 2EXAONE 4.0 32BOpenLG AI Research37.2$0.70——
32—Qwen3 4B (Reasoning)OpenAlibaba36.4$0.40——
33—Reka Flash 3OpenReka AI36.3$0.35——
34—Magistral SmallOpenMistral36.0$0.75——
35—DeepSeek V3 (Dec '24)OpenDeepSeek34.6$0.48——
36—Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)OpenNVIDIA34.1$0.000——
37—Qwen3 Coder 30B A3BOpenAlibaba33.4$0.90——
38—Qwen3 235B A22BOpenAlibaba33.3$1.2——
39—Llama 4 ScoutOpenMeta33.1$0.21——
40—Mistral Small 3.2OpenMistral31.8$0.15——
41—Command AOpenCohere31.6$4.4——
42—QwQ 32B-PreviewOpenAlibaba31.5$0.67——
43—Llama 3.3 Instruct 70BOpenMeta31.2$0.59——
44—Qwen3 30B A3BOpenAlibaba29.9$0.35——
45—Qwen3 14BOpenAlibaba29.8$0.61——
46—Qwen3 32BOpenAlibaba29.8$1.2——
47—Llama 3.1 Instruct 405BOpenMeta29.3$3.3——
48—Qwen2.5 Instruct 72BOpenAlibaba29.2$0.000——
49—MiniMax-Text-01OpenMiniMax29.1$0.43——
50—Llama 3.1 Tulu3 405BOpenAllen Institute for AI28.7$0.000——

Intelligence over time

Each maker's best model at the time on the Artificial Analysis Intelligence Index, by the month it was released.

Gained most · Anthropic +57.2Highest now · Anthropic 64.6Gap between highest and lowest 18.1 ptsTime range Nov 2025 – Nov 2026
64.6Anthropic60.0Google56.9OpenAI50.1Meta46.5Kimi

Arena data comes from the leaderboard dataset it publishes on Hugging Face (CC BY 4.0). Only each board's overall leaderboard is used; scores and ranks are unchanged. dataset · CC BY 4.0

Artificial Analysis data comes from its API (artificialanalysis.ai). Scores are unchanged; ranks follow the scores. artificialanalysis.ai

Open or closed follows the license Arena gives; prices and Artificial Analysis intelligence show only where model names match, otherwise “—”.