中文大模型能力榜

📊 全部資料 · 總覽

按 LMArena 中文競技場(Chinese Arena)的真人中文盲測對戰評分,排出大模型在中文場景的能力高下。它只統計中文對話裡的偏好勝負,和上方「總榜」(以英文為主)可能明顯不同——國產模型在這裡通常更靠前。評分反映偏好、非絕對能力,頭部分差常落在置信區間內(見下表),宜看作近似並列。資料歸因 LMArena,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	1552	±12	3.2K
2	claude-opus-4-6	anthropic	1548	±12	3.2K
3	gpt-5.5	openai	1539	±14	2.0K
4	qwen3.5-max-preview	alibaba	1537	±18	1.3K
5	claude-opus-4-7-thinking	anthropic	1536	±13	2.4K
6	gemini-3.5-flash-medium	google	1536	±28	462
7	claude-opus-4-7	anthropic	1532	±13	2.4K
8	kimi-k2.6	moonshot	1531	±15	1.8K
9	gpt-5.5-high	openai	1529	±14	2.1K
10	gemini-3-pro	google	1528	±14	2.1K
11	qwen3.7-max-preview	alibaba	1527	±37	267
12	gemini-3.1-pro-preview	google	1527	±11	4.1K
13	gpt-5.4-high	openai	1520	±12	2.9K
14	glm-5	zai	1517	±16	1.5K
15	gemini-3.5-flash-high	google	1517	±26	582
16	glm-5.1	zai	1515	±16	1.5K
17	qwen3.7-plus	alibaba	1513	±21	815
18	gemini-3-flash	google	1513	±15	1.5K
19	gpt-5.4	openai	1510	±12	3.1K
20	claude-opus-4-8	anthropic	1508	±18	1.3K
21	gemini-2.5-pro	google	1506	±8	6.5K
22	ernie-5.0-preview-1022	baidu	1506	±34	281
23	ernie-5.1	baidu	1505	±15	1.7K
24	mimo-v2.5-pro	xiaomi	1505	±15	1.9K
25	muse-spark	meta	1504	±21	854
26	qwen3.5-397b-a17b	alibaba	1502	±12	2.7K
27	glm-4.6	zai	1501	±15	1.6K
28	nvidia-nemotron-3-ultra-550b-a55b-nvfp4	nvidia	1500	±26	502
29	glm-5.2 (max)	zai	1499	±23	704
30	kimi-k2.5-thinking	moonshot	1499	±12	2.7K
31	ernie-5.0-0110	baidu	1499	±15	1.7K
32	claude-opus-4-8-thinking	anthropic	1496	±18	1.2K
33	deepseek-v4-pro	deepseek	1494	±14	2.0K
34	dola-seed-2.0-pro	bytedance	1493	±11	3.3K
35	gpt-5.1-high	openai	1493	±14	2.0K
36	claude-sonnet-5-high	anthropic	1492	±28	461
37	muse-spark-1.1	meta	1491	±42	186
38	qwen3.6-max-preview	alibaba	1487	±33	316
39	deepseek-v4-pro-thinking	deepseek	1486	±15	1.9K
40	qwen3-max-preview	alibaba	1486	±16	1.5K
41	claude-sonnet-4-6	anthropic	1486	±13	2.6K
42	glm-4.7	zai	1484	±22	687
43	grok-4.20-beta-0309-reasoning	xai	1481	±12	3.0K
44	ernie-5.0-preview-1203	baidu	1481	±25	536
45	grok-4.20-multi-agent-beta-0309	xai	1480	±12	2.8K
46	gemma-4-26b-a4b	google	1479	±29	388
47	grok-4.20-beta1	xai	1478	±17	1.4K
48	qwen3.6-plus	alibaba	1476	±15	1.9K
49	gpt-5.1	openai	1476	±13	2.0K
50	qwen3.5-27b	alibaba	1476	±16	1.5K
51	gemini-3-flash (thinking-minimal)	google	1476	±11	3.9K
52	grok-4-fast-chat	xai	1475	±32	356
53	mimo-v2-omni	xiaomi	1474	±21	864
54	deepseek-v4-flash-thinking	deepseek	1474	±14	2.0K
55	claude-opus-4-5-20251101	anthropic	1474	±11	3.7K
56	gemma-4-31b	google	1473	±30	374
57	grok-4.1	xai	1473	±11	3.7K
58	mimo-v2.5	xiaomi	1472	±15	1.8K
59	qwen3-next-80b-a3b-instruct	alibaba	1472	±17	1.2K
60	qwen3-235b-a22b-thinking-2507	alibaba	1472	±30	411
61	deepseek-v3.1-thinking	deepseek	1470	±21	874
62	amazon-nova-experimental-chat-11-10	amazon	1469	±17	1.2K
63	glm-4.5	zai	1469	±17	1.3K
64	longcat-flash-chat-2602-exp	meituan	1469	±15	1.6K
65	claude-sonnet-4-5-20250929-thinking-32k	anthropic	1467	±10	3.9K
66	qwen3.5-35b-a3b	alibaba	1466	±15	1.6K
67	qwen3-235b-a22b-instruct-2507	alibaba	1466	±9	5.0K
68	mimo-v2-flash (non-thinking)	xiaomi	1465	±13	2.3K
69	deepseek-v4-flash	deepseek	1465	±14	2.0K
70	claude-sonnet-4-5-20250929	anthropic	1464	±10	3.9K
71	minimax-m3	minimax	1464	±19	1.1K
72	qwen3.5-122b-a10b	alibaba	1464	±15	1.6K
73	deepseek-v3.2-exp	deepseek	1464	±23	624
74	claude-opus-4-5-20251101-thinking-32k	anthropic	1464	±14	1.8K
75	gpt-5.2-chat-latest-20260210	openai	1462	±15	1.8K
76	gemini-3.1-flash-lite-preview	google	1462	±11	3.3K
77	gpt-5.5-instant	openai	1461	±17	1.5K
78	kimi-k2.5-instant	moonshot	1461	±37	238
79	qwen3-vl-235b-a22b-instruct	alibaba	1460	±27	492
80	deepseek-v3.2	deepseek	1459	±13	2.4K
81	deepseek-r1-0528	deepseek	1459	±19	1.0K
82	deepseek-v3.1	deepseek	1458	±18	1.1K
83	deepseek-v3.2-thinking	deepseek	1457	±14	2.0K
84	amazon-nova-experimental-chat-26-02-10	amazon	1457	±38	216
85	deepseek-v3.2-exp-thinking	deepseek	1456	±33	304
86	qwen3.5-flash	alibaba	1456	±13	2.5K
87	mimo-v2-pro	xiaomi	1455	±18	1.3K
88	grok-4.1-thinking	xai	1454	±11	3.5K
89	kimi-k2-thinking-turbo	moonshot	1453	±11	3.2K
90	gpt-5.2-high	openai	1453	±12	2.5K
91	chatgpt-4o-latest-20250326	openai	1453	±9	4.1K
92	gemini-2.5-flash	google	1452	±8	6.4K
93	gemini-2.5-flash-preview-09-2025	google	1452	±16	1.4K
94	gpt-5.4-mini-high	openai	1452	±12	2.8K
95	qwen3-vl-235b-a22b-thinking	alibaba	1452	±35	296
96	gpt-5.2	openai	1450	±11	3.6K
97	amazon-nova-experimental-chat-12-10	amazon	1450	±35	238
98	mistral-medium-3.5	mistral	1450	±27	488
99	step-3.5-flash	stepfun	1449	±13	2.5K
100	grok-3-preview-02-24	xai	1448	±14	1.8K

能力評分資料來自 LMArena(原 LMSYS Chatbot Arena)公開發布的排行榜資料集「中文(chinese)」切片,許可 CC-BY-4.0,本站署名引用、未作修改。 Arena 分由真人匿名中文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。