Chatbot Arena (Text - Multi-Turn): leaderboard
The arena.ai text board restricted to battles whose conversation ran for more than one turn, so the vote covers an exchange rather than a single reply. Both models stay anonymous while the user keeps talking to them and then votes. The Bradley-Terry rating is refit over this subset of votes alone, and higher is better.
Metric: Arena Score. Source: arena.ai. 412 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon (High) | 1553 | #8 (Gemini 4 Argon) |
| 2 | Claude Opus 4.6 (High) | 1518 | #63 (Claude Opus 4.6) |
| 3 | Claude Opus 4.7 (High) | 1516 | #49 (Claude Opus 4.7) |
| 4 | Claude Fable 5 (High) | 1516 | #16 (Claude Fable 5) |
| 5 | Claude Opus 4.7 | 1514 | #49 |
| 6 | Claude Opus 4.6 | 1512 | #63 |
| 7 | Muse Spark 1.2 (xHigh) | 1506 | #62 (Muse Spark 1.2) |
| 8 | Gemini 3.8 Flash (High) | 1503 | #21 (Gemini 3.8 Flash) |
| 9 | Kimi K3 (Max) | 1500 | #30 (Kimi K3) |
| 10 | Claude Opus 5.5 (High) | 1498 | #3 (Claude Opus 5.5) |
| 11 | Gemini 3.1 Pro (Preview) | 1497 | #58 |
| 12 | Muse Spark 1.1 | 1497 | #52 |
| 13 | Claude Opus 4.8 (High) | 1497 | #37 (Claude Opus 4.8) |
| 14 | Gemini 3 Pro | 1496 | #80 |
| 15 | GPT-5.2 Instant | 1495 | #210 |
Interactive version: theaggregate.ai/benchmark?slug=chatbot-arena-text-multi-turn · How It Works · Data refreshed daily, snapshot 2026-10-20.