大模型能力榜
📊 全部数据 · 总览按 LMArena 竞技场(Chatbot Arena)的真人盲测对战评分,排出大模型的综合能力高下(overall 全球总榜)。Arena 分越高,代表在真人匿名对比中越受偏好——它反映的是偏好、非绝对能力,且头部几名的分差常落在置信区间内(见下表「置信区间」列),更宜看作近似并列。数据归因 LMArena,客观呈现、不构成投资或采购建议。榜单发布于 2026-10-15
| 排名 | 模型 | 机构 | Arena 分 | 置信区间 | 投票数 |
|---|---|---|---|---|---|
| 1 | Gemini 4 Argon (High) | 1532 | ±9 | 4.9K | |
| 2 | Claude Opus 5.5 (High) | anthropic | 1516 | ±7 | 8.3K |
| 3 | Claude Fable 5.1 (Max) | anthropic | 1512 | ±6 | 14.6K |
| 4 | Claude Opus 5 (Max) | anthropic | 1508 | ±5 | 32.5K |
| 5 | Claude Opus 5 (High) | anthropic | 1503 | ±4 | 66.8K |
| 6 | Claude Opus 4.6 (High) | anthropic | 1502 | ±3 | 81.4K |
| 7 | Muse Wonder (Max) | meta | 1501 | ±11 | 3.2K |
| 8 | Gemini 3.8 Flash (High) | 1499 | ±5 | 37.5K | |
| 9 | Claude Opus 4.6 | anthropic | 1498 | ±3 | 86.1K |
| 10 | Claude Fable 5 (High) | anthropic | 1491 | ±4 | 42.0K |
| 11 | mimo-v2.6-pro | xiaomi | 1491 | ±9 | 4.1K |
| 12 | Muse Spark 1.3 (Max) | meta | 1490 | ±6 | 17.6K |
| 13 | Claude Opus 4.7 (High) | anthropic | 1490 | ±4 | 65.1K |
| 14 | Gemini 3.7 Flash (High) | 1487 | ±5 | 27.2K | |
| 15 | Muse Spark 1.2 (xHigh) | meta | 1484 | ±7 | 8.2K |
| 16 | Claude Opus 4.7 | anthropic | 1483 | ±4 | 66.2K |
| 17 | qwen3.8-max | alibaba | 1483 | ±5 | 27.8K |
| 18 | gemini-3.5-flash-high | 1482 | ±4 | 48.5K | |
| 19 | Gemini 3.1 Pro Preview | 1480 | ±3 | 127.6K | |
| 20 | Gemini 3 Pro | 1479 | ±4 | 42.1K | |
| 21 | gemini-3.6-flash-high | 1479 | ±4 | 36.1K | |
| 22 | Muse Spark 1.1 | meta | 1478 | ±4 | 42.8K |
| 23 | gemini-3.5-flash-medium | 1477 | ±4 | 47.3K | |
| 24 | qwen3.7-max-preview | alibaba | 1475 | ±10 | 3.9K |
| 25 | Kimi K3 (Max) | moonshot | 1475 | ±5 | 32.8K |
| 26 | Muse Spark | meta | 1474 | ±6 | 14.2K |
| 27 | glm-5.3-max | zai | 1471 | ±5 | 22.5K |
| 28 | gpt-5.5-high | openai | 1471 | ±4 | 69.4K |
| 29 | glm-5.3-flash | zai | 1471 | ±5 | 32.9K |
| 30 | claude-sonnet-5.5-xhigh | anthropic | 1470 | ±8 | 6.8K |
| 31 | glm-5.2-max | zai | 1470 | ±4 | 49.8K |
| 32 | qwen3.5-max-preview | alibaba | 1470 | ±5 | 22.6K |
| 33 | gpt-5.4-high | openai | 1469 | ±4 | 64.8K |
| 34 | ernie-5.1 | baidu | 1468 | ±5 | 39.5K |
| 35 | gpt-5.5 | openai | 1467 | ±4 | 71.0K |
| 36 | gemini-3-flash | 1466 | ±4 | 31.3K | |
| 37 | mimo-v2.5-pro | xiaomi | 1465 | ±4 | 75.7K |
| 38 | deepseek-v4.1-flash-max | deepseek | 1463 | ±6 | 12.8K |
| 39 | glm-5.1 | zai | 1462 | ±4 | 62.9K |
| 40 | claude-opus-4-8-high | anthropic | 1461 | ±4 | 64.1K |
| 41 | claude-sonnet-4-6 | anthropic | 1458 | ±3 | 70.9K |
| 42 | gemini-2.5-pro | 1458 | ±2 | 125.4K | |
| 43 | gpt-5.6-sol-xhigh | openai | 1457 | ±4 | 40.4K |
| 44 | kimi-k2.6 | moonshot | 1455 | ±4 | 40.0K |
| 45 | mimo-v2.6-flash | xiaomi | 1455 | ±6 | 10.9K |
| 46 | qwen3.7-plus | alibaba | 1455 | ±4 | 42.1K |
| 47 | claude-opus-4-8 | anthropic | 1454 | ±4 | 65.2K |
| 48 | gpt-5.4 | openai | 1452 | ±4 | 67.9K |
| 49 | deepseek-v4-pro | deepseek | 1451 | ±4 | 57.6K |
| 50 | claude-opus-4-5-20251101 | anthropic | 1451 | ±3 | 72.9K |
| 51 | grok-4.20-beta-0309-reasoning | xai | 1451 | ±4 | 66.4K |
| 52 | grok-4.20-multi-agent-beta-0309 | xai | 1450 | ±4 | 64.9K |
| 53 | amazon-nova-experimental-chat-26-02-10 | amazon | 1448 | ±10 | 3.5K |
| 54 | claude-opus-4-5-20251101-high-32k | anthropic | 1448 | ±4 | 37.5K |
| 55 | dola-seed-2.0-pro | bytedance | 1448 | ±3 | 79.1K |
| 56 | grok-4.5 | xai | 1448 | ±4 | 39.9K |
| 57 | gpt-6.1-sol-max | openai | 1448 | ±8 | 6.3K |
| 58 | gpt-5.6-terra-xhigh | openai | 1447 | ±4 | 43.3K |
| 59 | qwen3.6-max-preview | alibaba | 1447 | ±8 | 5.4K |
| 60 | glm-5 | zai | 1446 | ±4 | 29.3K |
| 61 | deepseek-v4-pro-high-20260813 | deepseek | 1446 | ±7 | 9.8K |
| 62 | kimi-k2.5-thinking | moonshot | 1445 | ±3 | 74.8K |
| 63 | deepseek-v4-pro-high-preview | deepseek | 1445 | ±4 | 54.9K |
| 64 | nvidia-nemotron-3-ultra-550b-a55b-nvfp4 | nvidia | 1445 | ±7 | 11.6K |
| 65 | grok-4.20-beta1 | xai | 1445 | ±5 | 27.9K |
| 66 | ernie-5.0-0110 | baidu | 1445 | ±4 | 36.5K |
| 67 | gemma-4-31b | 1443 | ±7 | 6.1K | |
| 68 | gpt-5.1-high | openai | 1443 | ±4 | 41.7K |
| 69 | gpt-6-astra-max | openai | 1443 | ±6 | 12.2K |
| 70 | gemini-3-flash (thinking-minimal) | 1442 | ±3 | 91.0K | |
| 71 | ernie-5.0-preview-1203 | baidu | 1442 | ±6 | 9.8K |
| 72 | claude-sonnet-5-high | anthropic | 1442 | ±4 | 49.6K |
| 73 | Step 5 Preview | stepfun | 1441 | ±7 | 7.1K |
| 74 | inkling | thinky | 1441 | ±4 | 40.6K |
| 75 | qwen3.8-27b | alibaba | 1441 | ±5 | 23.3K |
| 76 | glm-4.6 | zai | 1440 | ±4 | 35.8K |
| 77 | qwen3-max-preview | alibaba | 1439 | ±5 | 27.4K |
| 78 | gpt-5.2-chat-latest-20260210 | openai | 1438 | ±4 | 36.0K |
| 79 | claude-sonnet-4-5-20250929 | anthropic | 1438 | ±3 | 82.7K |
| 80 | hy3 | tencent | 1438 | ±6 | 12.2K |
| 81 | qwen3.5-397b-a17b | alibaba | 1438 | ±3 | 93.6K |
| 82 | glm-5v-turbo | zai | 1438 | ±6 | 10.5K |
| 83 | grok-4.1-thinking | xai | 1436 | ±3 | 66.5K |
| 84 | qwen3.6-plus | alibaba | 1436 | ±4 | 48.2K |
| 85 | mimo-v2-pro | xiaomi | 1436 | ±5 | 25.4K |
| 86 | grok-4.1 | xai | 1436 | ±3 | 68.6K |
| 87 | glm-4.7 | zai | 1435 | ±6 | 12.3K |
| 88 | claude-sonnet-4-5-20250929-high-32k | anthropic | 1435 | ±3 | 83.9K |
| 89 | gemini-3.5-flash-lite | 1435 | ±4 | 35.9K | |
| 90 | gemma-4-26b-a4b | 1434 | ±7 | 6.1K | |
| 91 | minimax-m3 | minimax | 1433 | ±4 | 64.0K |
| 92 | deepseek-v4-flash | deepseek | 1432 | ±4 | 52.0K |
| 93 | gpt-5.6-luna-xhigh | openai | 1431 | ±4 | 41.1K |
| 94 | mistral-large-4 | mistral | 1430 | ±7 | 7.5K |
| 95 | glm-4.5 | zai | 1430 | ±5 | 23.8K |
| 96 | ernie-5.0-preview-1022 | baidu | 1430 | ±9 | 4.8K |
| 97 | chatgpt-4o-latest-20250326 | openai | 1428 | ±3 | 81.9K |
| 98 | mistral-large-3 | mistral | 1428 | ±3 | 84.2K |
| 99 | grok-4.6-high | xai | 1428 | ±5 | 29.0K |
| 100 | mimo-v2.5 | xiaomi | 1428 | ±4 | 47.4K |
中文能力评测参考
LMArena 总榜以英文对战为主;想看中文场景,可切到上方「中文榜」(LMArena 中文竞技场),或参考国内权威榜单(本站不转载其数据,仅外链):
能力评分数据来自 LMArena(原 LMSYS Chatbot Arena)公开发布的排行榜数据集,许可 CC-BY-4.0,本站署名引用、未作修改。 Arena 分由真人匿名盲测对战经统计模型(Bradley-Terry)估算,反映相对偏好、非绝对能力,可能滞后或随样本波动,仅供参考。 本栏目仅作信息整理,不构成任何投资或采购建议。本站不自建评测跑分。榜单发布于 2026-10-15。