大模型能力榜

📊 全部資料 · 總覽

按 LMArena 競技場(Chatbot Arena)的真人盲測對戰評分,排出大模型的綜合能力高下(overall 全球總榜)。Arena 分越高,代表在真人匿名對比中越受偏好——它反映的是偏好、非絕對能力,且頭部幾名的分差常落在置信區間內(見下表「置信區間」列),更宜看作近似並列。資料歸因 LMArena,客觀呈現、不構成投資或採購建議。榜單釋出於 2026-07-10

排名	模型	機構	Arena 分	置信區間	投票數
1	claude-opus-4-6-thinking	anthropic	1501	±4	59.0K
2	claude-opus-4-6	anthropic	1498	±4	62.6K
3	claude-fable-5	anthropic	1495	±9	4.3K
4	claude-opus-4-7-thinking	anthropic	1490	±4	46.2K
5	claude-opus-4-7	anthropic	1482	±4	47.2K
6	gemini-3.5-flash-high	google	1480	±6	10.1K
7	gemini-3-pro	google	1480	±4	41.3K
8	gemini-3.1-pro-preview	google	1479	±4	78.6K
9	muse-spark-1.1	meta	1478	±10	3.8K
10	gemini-3.5-flash-medium	google	1477	±7	9.5K
11	qwen3.7-max-preview	alibaba	1475	±10	3.7K
12	muse-spark	meta	1473	±6	13.6K
13	qwen3.5-max-preview	alibaba	1470	±5	21.5K
14	gpt-5.5-high	openai	1469	±5	41.3K
15	gpt-5.4-high	openai	1468	±4	54.6K
16	glm-5.1	zai	1467	±5	26.5K
17	ernie-5.1	baidu	1467	±5	36.3K
18	gemini-3-flash	google	1466	±4	30.7K
19	gpt-5.5	openai	1463	±4	42.6K
20	mimo-v2.5-pro	xiaomi	1462	±5	38.2K
21	claude-opus-4-8-thinking	anthropic	1461	±5	26.6K
22	glm-5.2 (max)	zai	1460	±6	13.4K
23	qwen3.7-plus	alibaba	1459	±6	18.5K
24	gpt-5.6-sol-xhigh	openai	1458	±14	1.7K
25	gemini-2.5-pro	google	1457	±2	124.5K
26	claude-sonnet-4-6	anthropic	1457	±4	52.9K
27	kimi-k2.6	moonshot	1456	±5	35.9K
28	grok-4.5	xai	1455	±11	3.2K
29	claude-opus-4-8	anthropic	1453	±5	27.0K
30	grok-4.20-beta-0309-reasoning	xai	1453	±4	55.9K
31	gpt-5.4	openai	1452	±4	57.3K
32	grok-4.20-multi-agent-beta-0309	xai	1450	±4	54.7K
33	claude-opus-4-5-20251101	anthropic	1449	±3	71.1K
34	deepseek-v4-pro	deepseek	1449	±4	41.0K
35	amazon-nova-experimental-chat-26-02-10	amazon	1448	±10	3.4K
36	dola-seed-2.0-pro	bytedance	1447	±4	63.5K
37	deepseek-v4-pro-thinking	deepseek	1447	±5	39.0K
38	claude-opus-4-5-20251101-thinking-32k	anthropic	1447	±4	37.1K
39	qwen3.6-max-preview	alibaba	1446	±8	5.2K
40	ernie-5.0-0110	baidu	1445	±4	35.3K
41	glm-5	zai	1445	±4	27.8K
42	kimi-k2.5-thinking	moonshot	1444	±4	59.1K
43	grok-4.20-beta1	xai	1444	±5	26.9K
44	gemini-3-flash (thinking-minimal)	google	1444	±3	79.9K
45	glm-5v-turbo	zai	1443	±10	3.4K
46	claude-sonnet-5-high	anthropic	1443	±7	9.0K
47	ernie-5.0-preview-1203	baidu	1442	±7	9.7K
48	gemma-4-31b	google	1442	±8	5.9K
49	gpt-5.1-high	openai	1441	±4	40.8K
50	nvidia-nemotron-3-ultra-550b-a55b-nvfp4	nvidia	1441	±7	9.4K
51	glm-4.6	zai	1440	±4	35.6K
52	qwen3-max-preview	alibaba	1439	±4	27.7K
53	gpt-5.2-chat-latest-20260210	openai	1438	±4	34.5K
54	qwen3.5-397b-a17b	alibaba	1438	±4	54.3K
55	claude-sonnet-4-5-20250929	anthropic	1438	±3	80.8K
56	grok-4.1-thinking	xai	1437	±3	65.5K
57	minimax-m3	minimax	1437	±6	23.8K
58	grok-4.1	xai	1437	±3	67.7K
59	qwen3.6-plus	alibaba	1436	±4	40.1K
60	glm-4.7	zai	1436	±6	12.1K
61	mimo-v2-pro	xiaomi	1436	±5	24.5K
62	gemma-4-26b-a4b	google	1434	±8	5.8K
63	deepseek-v4-flash	deepseek	1433	±4	40.8K
64	claude-sonnet-4-5-20250929-thinking-32k	anthropic	1433	±3	82.4K
65	mistral-large-3	mistral	1431	±3	47.1K
66	ernie-5.0-preview-1022	baidu	1430	±9	4.7K
67	glm-4.5	zai	1429	±5	24.3K
68	chatgpt-4o-latest-20250326	openai	1429	±3	82.4K
69	deepseek-r1-0528	deepseek	1428	±6	18.5K
70	mimo-v2.5	xiaomi	1427	±5	39.1K
71	longcat-flash-chat-2602-exp	meituan	1426	±5	28.1K
72	mistral-medium-2508	mistral	1425	±3	93.9K
73	grok-3-preview-02-24	xai	1425	±4	32.9K
74	deepseek-v3.2	deepseek	1424	±4	47.2K
75	deepseek-v4-flash-thinking	deepseek	1424	±5	40.6K
76	deepseek-v3.2-exp-thinking	deepseek	1424	±7	9.1K
77	deepseek-v3.2-exp	deepseek	1423	±6	11.9K
78	longcat-flash-chat	meituan	1422	±6	11.4K
79	gpt-5.1	openai	1422	±4	43.4K
80	mimo-v2-omni	xiaomi	1422	±6	19.5K
81	qwen3-vl-235b-a22b-instruct	alibaba	1421	±6	11.5K
82	mistral-medium-3.5	mistral	1421	±7	11.1K
83	kimi-k2.5-instant	moonshot	1421	±7	8.2K
84	deepseek-v3.2-thinking	deepseek	1420	±4	41.0K
85	deepseek-v3.1-terminus-thinking	deepseek	1420	±10	3.5K
86	amazon-nova-experimental-chat-12-10	amazon	1420	±10	3.7K
87	qwen3-235b-a22b-instruct-2507	alibaba	1419	±3	97.2K
88	gpt-5.5-instant	openai	1419	±5	26.1K
89	deepseek-v3.1	deepseek	1419	±6	14.9K
90	qwen3-next-80b-a3b-instruct	alibaba	1419	±5	22.9K
91	claude-opus-4-1-20250805-thinking-16k	anthropic	1418	±3	49.8K
92	qwen3.5-122b-a10b	alibaba	1418	±4	28.5K
93	claude-opus-4-1-20250805	anthropic	1417	±3	77.3K
94	gpt-4.5-preview-2025-02-27	openai	1417	±6	14.5K
95	deepseek-v3.1-terminus	deepseek	1417	±10	3.7K
96	gemini-2.5-flash	google	1417	±2	124.4K
97	deepseek-v3.1-thinking	deepseek	1417	±7	11.7K
98	amazon-nova-experimental-chat-11-10	amazon	1416	±4	25.4K
99	gpt-5.2-high	openai	1416	±4	48.0K
100	gemini-3.1-flash-lite-preview	google	1415	±4	61.0K

中文能力評測參考

LMArena 總榜以英文對戰為主;想看中文場景,可切到上方「中文榜」(LMArena 中文競技場),或參考國內權威榜單(本站不轉載其資料,僅外鏈):

能力評分資料來自 LMArena(原 LMSYS Chatbot Arena)公開發布的排行榜資料集,許可 CC-BY-4.0,本站署名引用、未作修改。 Arena 分由真人匿名盲測對戰經統計模型(Bradley-Terry)估算,反映相對偏好、非絕對能力,可能滯後或隨樣本波動,僅供參考。本欄目僅作資訊整理,不構成任何投資或採購建議。本站不自建評測跑分。榜單釋出於 2026-07-10。