Observed

实时·每 60 秒更新

模型排名

模型排名汇总

各来源标准不同，没有综合第一。故障请看状态页。

更新时间: 2026-08-06T09:32:36.710Z · 4/4 sources

Arena AI Text (community mirror)镜像来源怎么算截至: Aug 1, 2026

Arena Elo

People pick the better answer in blind A/B chats; that becomes Elo.

看什么: How chat and writing feel to people.

不看什么: Truthfulness, security, price, speed, or coding skill.

前列

Elo →

  1. 01
    Claude Fable 5anthropic
    1509 Elo
  2. 02
    Claude Opus 4 6 Thinkinganthropic
    1505 Elo
  3. 03
    Claude Opus 4 7 Thinkinganthropic
    1502 Elo
  4. 04
    Claude Opus 4 6anthropic
    1497 Elo
  5. 05
    Qwen3.8 Maxalibaba
    1496 Elo
  6. 06
    Claude Opus 4 7anthropic
    1492 Elo
  7. 07
    Claude Opus 5Highanthropic
    1492 Elo
  8. 08
    Claude Opus 5 Maxanthropic
    1490 Elo
  9. 09
    Muse Spark 1.1meta
    1490 Elo
  10. 10
    Muse Sparkmeta
    1488 Elo
  11. 11
    Gemini 3 Progoogle
    1486 Elo
  12. 12
    Gemini 3.1 Pro Previewgoogle
    1485 Elo
  13. 13
    Kimi K3 Maxmoonshot
    1485 Elo
  14. 14
    Claude Opus 4 8 Thinkinganthropic
    1484 Elo
  15. 15
    GPT 5.6 SolXHIGHopenai
    1483 Elo
  16. 16
    Gemini 3.6 Flashgoogle
    1483 Elo
  17. 17
    GPT 5.5Highopenai
    1482 Elo
  18. 18
    GPT 5.4Highopenai
    1477 Elo
  19. 19
    Gemini 3.5 FlashHighgoogle
    1476 Elo
  20. 20
    GPT 5.5openai
    1476 Elo

Bars are scaled to make gaps easier to see.

同一模型在各榜的名次。空着表示不在该榜前列。