---
title: "多轮对话排行榜"
description: "多轮对话的排名。"
canonical: https://pirouter.ai/zh/rankings/multiturn
last-updated: 2026-10-15
---

# 多轮对话排行榜

> 多轮对话的排名。

- 更新于: 2026-09-30
- 1,864,869 票 · 383 个模型
- 分数 · 越高越好

## 完整排名

| Rank | Rank spread | Model | Vendor | Score | Votes |
|---|---|---|---|---|---|
| 1 | 1–2 | Gemini 4 Argon (high) | Google | 1552 ±24 | 638 |
| 2 | 1–21 | Muse Spark 1.2 (xhigh) | Meta | 1522 ±24 | 603 |
| 3 | 2–6 | Claude Opus 4.6 (high) | Anthropic | 1519 ±6 | 13183 |
| 4 | 2–8 | Claude Fable 5 (high) | Anthropic | 1518 ±8 | 6231 |
| 5 | 2–8 | Claude Opus 4.7 (high) | Anthropic | 1516 ±7 | 11377 |
| 6 | 3–29 | [Gemini 3.8 Flash \(high\)](https://pirouter.ai/zh/models/google/gemini-3.8-flash) | Google | 1501 ±10 | 3796 |
| 7 | 3–29 | Kimi K3 (max) | Moonshot | 1500 ±10 | 4291 |
| 8 | 5–29 | Claude Opus 4.8 (high) | Anthropic | 1498 ±7 | 11116 |
| 9 | 2–66 | [Claude Opus 5.5 \(high\)](https://pirouter.ai/zh/models/anthropic/claude-opus-5.5) | Anthropic | 1497 ±26 | 496 |
| 10 | 5–29 | Gemini 3.1 Pro Preview | Google | 1497 ±5 | 20777 |
| 11 | 5–35 | Muse Spark 1.1 | Meta | 1496 ±9 | 5616 |
| 12 | 5–32 | Gemini 3 Pro | Google | 1496 ±8 | 6855 |
| 13 | 5–40 | [Gemini 3.7 Flash \(high\)](https://pirouter.ai/zh/models/google/gemini-3.7-flash) | Google | 1496 ±11 | 3167 |
| 14 | 5–36 | GPT-5.2 Chat Latest | OpenAI | 1495 ±8 | 6808 |
| 15 | 5–47 | Muse Spark | Meta | 1494 ±13 | 2284 |
| 16 | 5–37 | [GPT-5.4 \(high\)](https://pirouter.ai/zh/models/openai/gpt-5.4) | OpenAI | 1493 ±7 | 11980 |
| 17 | 5–56 | Muse Spark 1.3 (max) | Meta | 1491 ±14 | 1775 |
| 18 | 5–49 | Qwen3.8 (max) | Alibaba | 1490 ±10 | 3530 |
| 19 | 5–65 | [GPT-6 Astra \(max\)](https://pirouter.ai/zh/models/openai/gpt-6-astra) | OpenAI | 1489 ±17 | 1172 |
| 20 | 6–53 | [Gemini 3.6 Flash \(high\)](https://pirouter.ai/zh/models/google/gemini-3.6-flash) | Google | 1488 ±9 | 5704 |
| 21 | 6–45 | [GPT-5.5 \(high\)](https://pirouter.ai/zh/models/openai/gpt-5.5) | OpenAI | 1487 ±7 | 11837 |
| 22 | 5–65 | [Claude Fable 5.1 \(max\)](https://pirouter.ai/zh/models/anthropic/claude-fable-5.1) | Anthropic | 1487 ±16 | 1474 |
| 23 | 6–55 | [GPT-5.6 Sol \(xhigh\)](https://pirouter.ai/zh/models/openai/gpt-5.6-sol) | OpenAI | 1487 ±8 | 5924 |
| 24 | 6–56 | Claude Opus 5 (max) | Anthropic | 1486 ±10 | 4327 |
| 25 | 6–55 | Claude Opus 4.5 20251101 High 32k | Anthropic | 1486 ±8 | 6513 |
| 26 | 10–53 | Claude Opus 4.5 | Anthropic | 1484 ±6 | 12971 |
| 27 | 6–60 | Grok 4.20 Beta1 | xAI | 1483 ±9 | 4593 |
| 28 | 5–74 | Qwen3.7 Preview (max) | Alibaba | 1483 ±23 | 701 |
| 29 | 6–62 | GPT-5.5 Instant | OpenAI | 1482 ±9 | 5138 |
| 30 | 10–61 | Gemini 3 Flash | Google | 1482 ±9 | 5454 |
| 31 | 11–58 | Claude Sonnet 4.6 | Anthropic | 1482 ±6 | 12712 |
| 32 | 10–61 | [Gemini 3.5 Flash \(high\)](https://pirouter.ai/zh/models/google/gemini-3.5-flash) | Google | 1481 ±7 | 7997 |
| 33 | 12–60 | MiMo V2.5 Pro | Xiaomi | 1481 ±7 | 12065 |
| 34 | 6–67 | GLM 5.3 (max) | Z.ai | 1481 ±12 | 2623 |
| 35 | 14–63 | GLM 5.1 | Z.ai | 1479 ±7 | 9239 |
| 36 | 14–63 | Grok 4.20 Beta 0309 Reasoning | xAI | 1479 ±7 | 12129 |
| 37 | 15–62 | Claude Sonnet 4.5 | Anthropic | 1478 ±6 | 14479 |
| 38 | 11–67 | Qwen3.5 Preview (max) | Alibaba | 1477 ±10 | 3819 |
| 39 | 15–67 | [DeepSeek V4 Pro 0813](https://pirouter.ai/zh/models/deepseek/deepseek-v4-pro) | DeepSeek | 1475 ±7 | 10361 |
| 40 | 14–71 | [GLM 5.3 Flash](https://pirouter.ai/zh/models/z-ai/glm-5.3-flash) | Z.ai | 1475 ±11 | 3321 |
| 41 | 15–68 | [GLM 5.2 \(max\)](https://pirouter.ai/zh/models/z-ai/glm-5.2) | Z.ai | 1475 ±8 | 7241 |
| 42 | 16–67 | Grok 4.20 Multi Agent Beta 0309 | xAI | 1474 ±7 | 10624 |
| 43 | 18–67 | Gemini 3 Flash | Google | 1474 ±6 | 16312 |
| 44 | 15–69 | [Claude Sonnet 5 \(high\)](https://pirouter.ai/zh/models/anthropic/claude-sonnet-5) | Anthropic | 1474 ±8 | 7677 |
| 45 | 16–71 | Grok 4.5 | xAI | 1472 ±8 | 6568 |
| 46 | 22–68 | Claude Sonnet 4.5 20250929 High 32k | Anthropic | 1472 ±6 | 14320 |
| 47 | 17–71 | Ernie 5.1 | Baidu | 1472 ±8 | 6816 |
| 48 | 20–71 | Claude Opus 4.1 (thinking) | Anthropic | 1471 ±7 | 8467 |
| 49 | 17–71 | GLM 5 | Z.ai | 1471 ±9 | 4698 |
| 50 | 24–71 | Chatgpt 4o Latest | OpenAI | 1471 ±6 | 14281 |
| 51 | 20–72 | [GPT-5.6 Terra \(xhigh\)](https://pirouter.ai/zh/models/openai/gpt-5.6-terra) | OpenAI | 1470 ±8 | 6061 |
| 52 | 13–88 | [GPT-6 Sol \(max\)](https://pirouter.ai/zh/models/openai/gpt-6-sol) | OpenAI | 1469 ±18 | 1148 |
| 53 | 11–92 | Qwen3.6 Preview (max) | Alibaba | 1468 ±20 | 932 |
| 54 | 24–75 | GPT-5.3 Chat Latest | OpenAI | 1467 ±8 | 6456 |
| 55 | 18–86 | GPT-4.5 Preview 2025 02 27 | OpenAI | 1466 ±13 | 2052 |
| 56 | 31–75 | [GPT-5.4 Mini \(high\)](https://pirouter.ai/zh/models/openai/gpt-5.4-mini) | OpenAI | 1466 ±7 | 11875 |
| 57 | 15–92 | Gemma 4 31b | Google | 1465 ±17 | 1128 |
| 58 | 25–79 | MiMo V2 Pro | Xiaomi | 1465 ±9 | 4601 |
| 59 | 29–79 | [Gemini 3.5 Flash Lite](https://pirouter.ai/zh/models/google/gemini-3.5-flash-lite) | Google | 1465 ±9 | 5773 |
| 60 | 33–79 | GPT-5.1 (high) | OpenAI | 1464 ±8 | 7243 |
| 61 | 33–79 | Qwen3.7 Plus | Alibaba | 1464 ±8 | 7017 |
| 62 | 18–96 | [DeepSeek V4.1 Flash \(max\)](https://pirouter.ai/zh/models/deepseek/deepseek-v4.1-flash) | DeepSeek | 1462 ±17 | 1178 |
| 63 | 36–85 | Kimi K2.6 | Moonshot | 1461 ±8 | 6944 |
| 64 | 41–82 | Grok 4.1 | xAI | 1461 ±6 | 12432 |
| 65 | 35–96 | GLM 4.7 | Z.ai | 1457 ±14 | 1929 |
| 66 | 44–91 | [GPT-5.6 Luna \(xhigh\)](https://pirouter.ai/zh/models/openai/gpt-5.6-luna) | OpenAI | 1457 ±8 | 6086 |
| 67 | 18–111 | MiMo V2.6 Pro | Xiaomi | 1455 ±24 | 580 |
| 68 | 27–107 | [GPT-6 Luna \(max\)](https://pirouter.ai/zh/models/openai/gpt-6-luna) | OpenAI | 1455 ±18 | 1115 |
| 69 | 52–92 | Qwen3.5 397b A17b | Alibaba | 1454 ±6 | 14408 |
| 70 | 25–111 | MiMo V2.6 Flash | Xiaomi | 1453 ±21 | 818 |
| 71 | 51–95 | [DeepSeek V4 Flash 0731](https://pirouter.ai/zh/models/deepseek/deepseek-v4-flash) | DeepSeek | 1453 ±7 | 9354 |
| 72 | 54–95 | Kimi K2.5 (thinking) | Moonshot | 1452 ±6 | 13156 |
| 73 | 50–102 | MiMo V2 Omni | Xiaomi | 1452 ±11 | 3508 |
| 74 | 43–108 | Hy3 | Tencent | 1451 ±15 | 1658 |
| 75 | 54–96 | MiMo V2.5 | Xiaomi | 1451 ±8 | 8378 |
| 76 | 54–96 | Minimax M3 | MiniMax | 1450 ±7 | 10225 |
| 77 | 54–99 | GPT-5 Chat | OpenAI | 1449 ±8 | 5684 |
| 78 | 44–111 | Gemma 4 26b A4b | Google | 1449 ±17 | 1145 |
| 79 | 58–97 | Dola Seed 2.0 Pro | ByteDance | 1449 ±6 | 13822 |
| 80 | 60–96 | Gemini 2.5 Pro | Google | 1448 ±5 | 21192 |
| 81 | 59–102 | Grok 4.3 | xAI | 1447 ±7 | 12878 |
| 82 | 51–111 | Qwen3 Max 2025 09 23 | Alibaba | 1447 ±15 | 1636 |
| 83 | 61–107 | Qwen3.6 Plus | Alibaba | 1445 ±8 | 8438 |
| 84 | 62–105 | GPT-5.2 | OpenAI | 1445 ±6 | 15952 |
| 85 | 59–109 | Grok 4.6 (high) | xAI | 1444 ±11 | 3507 |
| 86 | 59–111 | Qwen3.8 27b | Alibaba | 1443 ±11 | 2877 |
| 87 | 61–109 | Qwen3 Preview (max) | Alibaba | 1442 ±9 | 4526 |
| 88 | 62–109 | Ernie 5.0 0110 | Baidu | 1442 ±8 | 5910 |
| 89 | 58–121 | Kimi K2.5 Instant | Moonshot | 1440 ±15 | 1499 |
| 90 | 63–111 | Inkling | Thinky | 1440 ±9 | 5491 |
| 91 | 66–111 | Claude Opus 4 (thinking) | Anthropic | 1439 ±8 | 6044 |
| 92 | 74–111 | Qwen3 235b A22b Instruct 2507 | Alibaba | 1436 ±5 | 17061 |
| 93 | 74–115 | Gemini 3.1 Flash Lite Preview | Google | 1436 ±7 | 11615 |
| 94 | 66–127 | Mistral Medium 3.5 | Mistral | 1433 ±14 | 1924 |
| 95 | 77–120 | Kimi K2 Thinking Turbo | Moonshot | 1433 ±6 | 10984 |
| 96 | 66–129 | GLM 5v Turbo | Z.ai | 1433 ±15 | 1730 |
| 97 | 75–122 | Longcat Flash Chat 2602 Exp | Meituan | 1432 ±9 | 5248 |
| 98 | 58–145 | Muse Glimmer | Meta | 1431 ±24 | 625 |
| 99 | 80–122 | GPT-4.1 2025 04 14 | OpenAI | 1430 ±7 | 8731 |
| 100 | 81–124 | DeepSeek V3.2 | DeepSeek | 1429 ±7 | 8337 |

## FAQ

### 多轮对话最好的 AI 模型是哪个？

截至 2026-09-30，多轮对话榜第一为 Google Gemini 4 Argon（high），分数 1552，名次区间 1–2。

### 多轮对话最好的开放权重模型是哪个？

Kimi K3（Moonshot），全榜第 7 名。

### 这个榜单排的是什么？

多轮对话的排名。

### 多久更新一次？

有新模型发布或排名变化时更新；本榜最近一次更新为 2026-09-30。

数据：[LMArena leaderboard dataset](https://huggingface.co/datasets/lmarena-ai/leaderboard-dataset)，按 [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) 许可使用；仅做筛选与推理档位合并，分数未改动。
查看原榜单: https://arena.ai/leaderboard/text/multi-turn