LMArena Multi-turn
Text arena rating on conversations with more than one turn.
As of 19 Sept 2026, Muse Spark 1.2 leads LMArena Multi-turn on BenchLeader with 1520, ahead of Claude Fable 5 at 1518, across 364 model configurations with a published result.
- Published by
- LMArena
- Category
- Human preference
- Index weight
- Reference only
- Models
- 364
- Data as of
- 19 Sept 2026
CC BY 4.0 — lmarena-ai/leaderboard-dataset on Hugging Face.
What the test looks like
Pairwise votes on multi-turn conversations, rewarding models that track context.
How it is scored
Bradley-Terry rating from pairwise votes, in Elo-like units, with style control.
What to keep in mind
Fewer votes than the overall board.
- 1Muse Spark 1.2 (xhigh)1520
- 2Claude Fable 51518
- 3Claude Opus 4.6 (high)1517
- 4Claude Opus 4.7 (high)1516
- 5Claude Opus 4.71515
- 6Gemini 3.8 Flash (high)1513
- 7Claude Opus 4.61511
- 8GPT-6 Astra (max)1499
- 9Gemini 3.7 Flash (high)1498
- 10Muse Spark 1.11497
- 11Claude Opus 4.8 (high)1496
- 12Claude Opus 4.81496
- 13Kimi K3 (max)1496
- 14Gemini 3 Pro1495
- 15Gemini 3.1 Pro1495
364 of 364
| # | ||||
|---|---|---|---|---|
| 1 |  Muse Spark 1.2xhighMeta | 1520 | 64.1 | 2026-09-13 |
| 2 |  Claude Fable 5Anthropic | 1518 | 68.3 | 2026-09-13 |
| 3 | Claude Opus 4.6highAnthropic | 1517 | 61.3 | 2026-09-13 |
| 4 | Claude Opus 4.7highAnthropic | 1516 | 62.4 | 2026-09-13 |
| 5 | Claude Opus 4.7Anthropic | 1515 | 64.5 | 2026-09-13 |
| 6 |  Gemini 3.8 FlashhighGoogle | 1513 | 64.5 | 2026-09-13 |
| 7 | Claude Opus 4.6Anthropic | 1511 | 63.7 | 2026-09-13 |
| 8 |  GPT-6 AstramaxOpenAI | 1499 | 71.8 | 2026-09-13 |
| 9 | Gemini 3.7 FlashhighGoogle | 1498 | 64.6 | 2026-09-13 |
| 10 | Muse Spark 1.1Meta | 1497 | 65.2 | 2026-09-13 |
| 11 | Claude Opus 4.8highAnthropic | 1496 | 62.2 | 2026-09-13 |
| 12 | Claude Opus 4.8Anthropic | 1496 | 61.9 | 2026-09-13 |
| 13 |  Kimi K3maxMoonshot AIopen ↗ | 1496 | 67.2 | 2026-09-13 |
| 14 | Gemini 3 ProGoogle | 1495 | 61.1 | 2026-09-13 |
| 15 | Gemini 3.1 ProGoogle | 1495 | 63.9 | 2026-09-13 |
| 16 |  Qwen3 8maxAlibaba | 1495 | 66.5 | 2026-09-13 |
| 17 | GPT-5.4highOpenAI | 1494 | 59.0 | 2026-09-13 |
| 18 | GPT-5.2OpenAI | 1494 | 58.4 | 2026-09-13 |
| 19 | Muse Spark 1.3maxMeta | 1493 | 69.3 | 2026-09-13 |
| 20 |  GLM-5.3maxZhipu AIopen ↗ | 1493 | 65.8 | 2026-09-13 |
| 21 | Muse SparkMeta | 1491 | 65.7 | 2026-09-13 |
| 22 | GPT-5.5highOpenAI | 1489 | 67.0 | 2026-09-13 |
| 23 | Claude Fable 5.1maxAnthropic | 1488 | 69.7 | 2026-09-13 |
| 24 | GPT-5.6 SolxhighOpenAI | 1487 | 68.2 | 2026-09-13 |
| 25 | Claude Opus 4.5highAnthropic | 1486 | 57.7 | 2026-09-13 |
| 26 | Gemini 3.6 FlashhighGoogle | 1485 | 61.7 | 2026-09-13 |
| 27 | Claude Opus 5maxAnthropic | 1485 | 69.9 | 2026-09-13 |
| 28 | Claude Opus 4.5Anthropic | 1484 | 58.2 | 2026-09-13 |
| 29 | Claude Opus 5highAnthropic | 1484 | 70.2 | 2026-09-13 |
| 30 | Gemini 3.5 FlashhighGoogle | 1483 | 63.6 | 2026-09-13 |
| 31 | GPT-5.5 InstantOpenAI | 1483 | 59.2 | 2026-09-13 |
| 32 | Gemini 3 FlashGoogle | 1483 | 57.7 | 2026-09-13 |
| 33 |  Grok 4.20 Beta1xAI | 1482 | – | 2026-09-13 |
| 34 | Qwen3 7maxAlibaba | 1482 | 61.0 | 2026-09-13 |
| 35 | GPT-5.5OpenAI | 1481 | 63.2 | 2026-09-13 |
| 36 | Claude Sonnet 4.6Anthropic | 1481 | 59.0 | 2026-09-13 |
| 37 | GPT-5.4OpenAI | 1481 | 59.2 | 2026-09-13 |
| 38 | Grok 4.20thinkingxAI | 1480 | 59.9 | 2026-09-13 |
| 39 | Qwen3 5maxAlibaba | 1479 | – | 2026-09-13 |
| 40 | Claude Sonnet 4.5Anthropic | 1478 | 54.3 | 2026-09-13 |
| 41 |  MiMo-V2.5-ProXiaomiopen ↗ | 1478 | 59.2 | 2026-09-13 |
| 42 | GLM-5.1Zhipu AIopen ↗ | 1477 | 56.9 | 2026-09-13 |
| 43 | Grok 4.5xAI | 1476 | 60.3 | 2026-09-13 |
| 44 | Gemini 3.5 FlashmediumGoogle | 1476 | 64.2 | 2026-09-13 |
| 45 |  DeepSeek V4 ProDeepSeekopen ↗ | 1475 | 52.6 | 2026-09-13 |
| 46 | GLM-5.3-FlashZhipu AIopen ↗ | 1475 | 63.9 | 2026-09-13 |
| 47 | Gemini 3 FlashminimalGoogle | 1474 | 54.9 | 2026-09-13 |
| 48 | Claude Sonnet 5highAnthropic | 1473 | 62.2 | 2026-09-13 |
| 49 | Grok 4.20 Multi-AgentxAI | 1473 | 59.2 | 2026-09-13 |
| 50 | GPT-5.6 TerraxhighOpenAI | 1473 | 64.3 | 2026-09-13 |
| 51 | GLM-5Zhipu AIopen | 1473 | 54.7 | 2026-09-13 |
| 52 | Claude Opus 4.1thinkingAnthropic | 1473 | 57.2 | 2026-09-13 |
| 53 |  Ernie 5.1Baidu | 1473 | – | 2026-09-13 |
| 54 | GPT-4oOpenAI | 1471 | 44.3 | 2026-09-13 |
| 55 | Claude Sonnet 4.5highAnthropic | 1470 | 58.8 | 2026-09-13 |
| 56 | Deepseek v4 ProhighDeepSeek | 1469 | 60.8 | 2026-09-13 |
| 57 | GLM-5.2maxZhipu AIopen ↗ | 1469 | 63.9 | 2026-09-13 |
| 58 | Qwen3 6maxAlibaba | 1469 | 63.3 | 2026-09-13 |
| 59 | Mimo v2 ProXiaomi | 1469 | 59.8 | 2026-09-13 |
| 60 | Claude Opus 4.1Anthropic | 1469 | 51.9 | 2026-09-13 |
Cite as: BenchLeader, “LMArena Multi-turn leaderboard”, https://www.benchleader.com/benchmarks/lmarena_text_multiturn, data as of 19 Sept 2026.
LMArena Multi-turn: questions
- What does LMArena Multi-turn measure?
- Pairwise votes on multi-turn conversations, rewarding models that track context. Scores are reported in Elo-style ratings from pairwise comparisons; higher is better.
- Which AI model leads LMArena Multi-turn?
- Muse Spark 1.2 leads LMArena Multi-turn with 1520 as of 19 Sept 2026, ahead of Claude Fable 5 at 1518.
- How many models have LMArena Multi-turn results?
- 364 model configurations have a LMArena Multi-turn result on BenchLeader, all taken from LMArena.
- Who runs LMArena Multi-turn and how often is it updated?
- LMArena Multi-turn is published by LMArena. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does LMArena Multi-turn count toward the BenchLeader Index?
- No. LMArena Multi-turn is shown for reference but left out of the composite index.