Skip to content
IFAI

Model rankings

994 models · 5 boards · 2 sources

Artificial AnalysisIntelligence IndexArenaText arena

Intelligence Index · A composite of several evaluations · Updated February 5, 2026

Smartest
GPT-5.2 (xhigh)OpenAI51.2Intelligence
Best open model
Kimi K2.5 (Reasoning)Kimi46.7Intelligence
Best value
MiMo-V2-Flash (Reasoning)Xiaomi262 pts/$Value
Best from China
Kimi K2.5 (Reasoning)Kimi46.7Intelligence
Top three by Intelligence
1GPT-5.2 (xhigh)OpenAI51.2
Top three open models
1Kimi K2.5 (Reasoning)Kimi · #6 overall46.7
2GLM-4.7 (Reasoning)Z AI · #12 overall42.1
3DeepSeek V3.2 (Reasoning)DeepSeek · #13 overall41.6
This week's changes
claude-opus-4-6Coding · #1New
Kimi K2.5 (Reasoning)Intelligence · #6New
kimi-k2.5-instantCoding · #9New
Qwen3-Coder-NextIntelligence · #53New
Step3 VL 10BIntelligence · #151New
RankChangeModelIntelligencePrice $/MSpeed tok/sFirst token
1—GPT-5.2 (xhigh)ClosedOpenAI51.2$4.8——
2—Claude Opus 4.5 (Reasoning)ClosedAnthropic49.7$10——
3—GPT-5.2 Codex (xhigh)ClosedOpenAI49.0$4.8——
4—Gemini 3 Pro Preview (high)ClosedGoogle48.4$4.5——
5—GPT-5.1 (high)ClosedOpenAI47.6$3.4——
6—GPT-5.2 (medium)ClosedOpenAI46.6$4.8——
7—Gemini 3 Flash Preview (Reasoning)ClosedGoogle46.4$1.1——
8—GPT-5 (high)ClosedOpenAI44.6$3.4——
9—GPT-5 Codex (high)ClosedOpenAI44.5$3.4——
10—Claude Opus 4.5 (Non-reasoning)ClosedAnthropic43.1$10——
11—Claude 4.5 Sonnet (Reasoning)ClosedAnthropic42.9$6.0——
12—GPT-5.1 Codex (high)ClosedOpenAI42.2$3.4——
13—GPT-5 (medium)ClosedOpenAI41.8$3.4——
14—Grok 4ClosedxAI41.4$6.0——
15—Gemini 3 Pro Preview (low)ClosedGoogle41.1$4.5——
16—GPT-5 mini (high)ClosedOpenAI41.0$0.69——
17—o3ClosedOpenAI40.9$3.5——
18—o3-proClosedOpenAI40.7$35——
19—GPT-5 (low)ClosedOpenAI39.0$3.4——
20—GPT-5 mini (medium)ClosedOpenAI38.8$0.69——
21—Claude 4 Sonnet (Reasoning)ClosedAnthropic38.6$6.0——
22—GPT-5.1 Codex mini (high)ClosedOpenAI38.5$0.69——
22—Grok 4.1 Fast (Reasoning)ClosedxAI38.5$0.28——
24—Claude 4.5 Sonnet (Non-reasoning)ClosedAnthropic37.1$6.0——
25—Claude 4.5 Haiku (Reasoning)ClosedAnthropic37.0$2.0——
26—Nova 2.0 Pro Preview (medium)ClosedAmazon35.6$3.4——
27—Gemini 3 Flash Preview (Non-reasoning)ClosedGoogle35.1$1.1——
28—Grok 4 Fast (Reasoning)ClosedxAI34.9$0.28——
29—Claude 3.7 Sonnet (Reasoning)ClosedAnthropic34.6$6.0——
30—Gemini 2.5 ProClosedGoogle34.5$3.4——
31—GPT-5.2 (Non-reasoning)ClosedOpenAI33.5$4.8——
32—gpt-oss-120B (high)OpenOpenAI33.3$0.26——
33—o4-mini (high)ClosedOpenAI33.1$1.9——
34—Claude 4 Sonnet (Non-reasoning)ClosedAnthropic33.0$6.0——
35—Grok 3 mini Reasoning (high)ClosedxAI32.0$0.35——
36—Claude 4.1 Opus (Reasoning)ClosedAnthropic31.9$30——
37—Nova 2.0 Pro Preview (low)ClosedAmazon31.9$3.4——
38—Gemini 2.5 Flash Preview (Sep '25) (Reasoning)ClosedGoogle31.1$0.85——
39—Claude 4.5 Haiku (Non-reasoning)ClosedAnthropic31.0$2.0——
40—Claude 3.7 Sonnet (Non-reasoning)ClosedAnthropic30.8$6.0——
41—o1ClosedOpenAI30.7$26——
42—Gemini 2.5 Pro Preview (Mar' 25)ClosedGoogle30.3$0.000——
43—Nova 2.0 Lite (medium)ClosedAmazon29.7$0.85——
44—Gemini 2.5 Pro Preview (May' 25)ClosedGoogle29.5$3.4——
45—Grok Code Fast 1ClosedxAI28.7$0.53——
46—Apriel-v1.5-15B-ThinkerOpenServiceNow28.3$0.000——
47—Nova 2.0 Omni (medium)ClosedAmazon27.9$0.85——
48—Apriel-v1.6-15B-ThinkerOpenServiceNow27.5$0.000——
49—GPT-5.1 (Non-reasoning)ClosedOpenAI27.4$3.4——
50—Claude 4 Opus (Reasoning)ClosedAnthropic27.4$30——

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 “—”.