Leaderboard
The top 10 on September 23, 2025
Ranked by the Artificial Analysis Intelligence Index, with each model's price on that day (per million tokens, 3 parts input to 1 part output).
- 1GPT-5 CodexClosedOpenAI · released Sep 23, 2025 · out that dayIntelligence 24.9$3.44 /M
- 2GPT-5ClosedOpenAI · released Aug 7, 2025 · 7 weeks oldIntelligence 23$3.44 /M
- 3Claude 4.1 OpusClosedAnthropic · released Aug 5, 2025 · 7 weeks oldIntelligence 22.8$30 /M
- 4Grok 4ClosedxAI · released Jul 10, 2025 · 2 months oldIntelligence 22.5$6 /M
- 5o3-proClosedOpenAI · released Jun 10, 2025 · 3 months oldIntelligence 21.9$35 /M
- 6Claude 4 OpusClosedAnthropic · released May 22, 2025 · 4 months oldIntelligence 20.6$30 /M
- 7GPT-5 miniClosedOpenAI · released Aug 7, 2025 · 7 weeks oldIntelligence 20.6$0.69 /M
- 8o3ClosedOpenAI · released Apr 16, 2025 · 5 months oldIntelligence 20.2$3.50 /M
- 9Claude 4 SonnetClosedAnthropic · released May 22, 2025 · 4 months oldIntelligence 18.9$6 /M
- 10Grok 4 FastClosedxAI · released Sep 19, 2025 · 4 days oldIntelligence 17.9$0.28 /M
13 months later
What has happened since
260 models have come out since. Today's best is Claude Opus 5.5 at 57.6, 32.7 points higher.
Agnes 2.5 Pro Beta scores at least as well today for $0.15 per million tokens, against $3.44 for GPT-5 Codex back then.
DeepSeek V4 Pro 0424 matched it for $0.54 per million tokens on May 22, 2026.
GLM-5 from Z.ai matched it on Feb 11, 2026, free to download and run.
The race
Which lab held the top spot
OpenAI led for 2.5 years of the 4 years since ChatGPT launched. Anthropic has held #1 since Apr 16, 2026, 6 models in a row. We rank by the Artificial Analysis Intelligence Index. Other leaderboards, such as LMArena's human votes, crown different models.
The wider field
240 models were out on Sep 23, 2025
Kimi, scoring 15.3. #15 overall, 9.6 points behind.
OpenAI, scoring 20.6. #7 overall at $0.69, 5× cheaper than #1.
- GPT-5 Codex24.9
- Qwen3 Max15.6
- Qwen3 VL 235B A22B13.4
- Qwen3 VL 235B A22B Instruct9.9
- LFM2 2.6B5.2
- DeepSeek V3.1 Terminus14.8
Browse
The top AI model, month by month
Try today's top models on your own prompt
Scores tell you who leads on average. Send one prompt to up to six models and see which answer you'd actually use, with what each one cost. For the long view of prices, read Six months later, the best AI costs 20× less.
Scores and release dates from Artificial Analysis (Intelligence Index, best published effort of each model), data through October 18, 2026. Prices are list prices on the day, including later cuts we know of. Models without a public price are ranked but not priced.