LMArena is a widely referenced platform that ranks large AI models across multiple areas—including text, coding, vision, and specialized tasks—using extensive human preference data. Its leaderboards reflect which models users find most capable in real-world tasks, based on head-to-head comparisons and live voting.
LMArena assesses AI models in several distinct arenas:
Text: General language tasks (e.g., writing, reasoning, multi-turn dialogue)
WebDev: Coding and web development capabilities
Vision: Multimodal ability to handle images and text
Search: Information retrieval and synthesis
Copilot: Coding assistant and developer support
Text-to-Image: AI systems generating visual content from text prompts
Current Rankings by Area
Here are the current (July 2025) rankings.
Text Arena
Rank
Model
Score
Votes
1
Gemini-2.5-Pro
1462
18,297
2
o3-2025-04-16
1452
24,554
2
ChatGPT-4o-Latest-20250326
1444
25,715
3
GPT-4.5-Preview-2025-02-27
1437
15,271
3
Grok-4-0709
1433
4,227
5
Claude-Opus-4-20250514-Thinking-16k
1419
13,018
6
Claude-Opus-4-20250514
1416
21,129
6
DeepSeek-R1-0528
1414
14,078
6
Gemini-2.5-Flash
1414
23,738
6
GPT-4.1-2025-04-14
1412
19,766
WebDev Arena
Rank
Model
Score
Votes
1
Gemini-2.5-Pro
1423
3,010
1
DeepSeek-R1-0528
1407
1,978
1
Claude Opus 4 (20250514)
1404
4,322
3
Claude Sonnet 4 (20250514)
1378
3,258
4
Claude 3.7 Sonnet (20250219)
1357
7,481
6
Gemini-2.5-Flash
1299
3,681
Vision Arena
Rank
Model
Score
Votes
1
Gemini-2.5-Pro
1268
4,382
2
ChatGPT-4o-Latest-20250326
1249
6,271
2
o3-2025-04-16
1238
5,097
2
GPT-4.5-Preview-2025-02-27
1231
3,066
3
Gemini-2.5-Flash
1224
5,184
Search Arena
Rank
Model
Score
Votes
1
Gemini-2.5-Pro-Grounding
1142
1,215
1
PPL-Sonar-Reasoning-Pro-High
1136
861
3
PPL-Sonar-Reasoning
1097
1,644
3
PPL-Sonar
1072
1,208
Copilot Arena
Rank
Model
Score
Votes
1
DeepSeek V2.5 (FIM)
1028
2,292
1
Claude 3.5 Sonnet (06/20)
1012
3,544
1
Claude 3.5 Sonnet (10/22)
1004
3,596
1
Codestral (25.01)
1001
2,180
1
Qwen-2.5-Coder (FiM)
998
3,401
Text-to-Image Arena
Rank
Model
Score
Votes
1
GPT-Image-1
1148
22,691
2
Imagen-4.0-Ultra-Generate-Preview-06-06
1113
11,552
3
Imagen-4.0-Generate-Preview-05-20
1097
22,211
And the winner is…
Clearly, Gemini-2.5-Pro leads most areas, especially in Text, Vision, and WebDev. Google, again…
