模型排行

排名	变化	模型	得分	置信区间	投票	价格 $/M
1	—	qwen3-235b-a22b-instruct-2507开源Alibaba	1,419	±5	21,971	—
2	—	kimi-k2-0711-preview开源Moonshot AI · 另有 1 个档位	1,418	±5	26,261	—
3	—	deepseek-r1-0528开源DeepSeek · 另有 1 个档位	1,418	±6	20,358	—
4	—	deepseek-v3.1开源DeepSeek · 另有 1 个档位	1,417	±6	13,851	—
5	—	glm-4.5开源Z.ai	1,408	±6	16,882	—
6	—	qwen3-next-80b-a3b-instruct开源Alibaba	1,400	±8	6,103	—
7	—	qwen3-235b-a22b-no-thinking开源Alibaba	1,400	±5	35,755	—
8	—	qwen3-235b-a22b-thinking-2507开源Alibaba	1,398	±7	8,293	—
9	—	longcat-flash-chat开源Meituan	1,397	±9	4,937	—
10	—	deepseek-v3-0324开源DeepSeek · 另有 1 个档位	1,390	±4	39,009	—
11	—	qwen3-30b-a3b-instruct-2507开源Alibaba	1,380	±6	15,480	—
12	—	qwen3-coder-480b-a35b-instruct开源Alibaba	1,379	±5	18,254	—
13	—	qwen3-next-80b-a3b-thinking开源Alibaba	1,373	±8	5,443	—
14	—	qwen3-235b-a22b开源Alibaba	1,373	±5	24,561	—
15	—	glm-4.5-air开源Z.ai	1,372	±6	15,986	—
16	—	minimax-m1开源MiniMax	1,369	±5	27,203	—
17	—	gemma-3-27b-it开源Google	1,363	±4	39,569	—
18	—	mistral-small-2506开源Mistral	1,351	±6	15,088	—
19	—	gpt-oss-120b开源OpenAI	1,350	±6	14,753	$0.26
20	—	step-3开源StepFun	1,349	±11	3,073	—
21	—	command-a-03-2025开源Cohere	1,347	±4	42,057	—
22	—	llama-3.1-nemotron-ultra-253b-v1开源NVIDIA	1,344	±12	2,573	—
23	—	qwen3-32b开源Alibaba	1,343	±10	3,943	—
24	—	nvidia-llama-3.3-nemotron-super-49b-v1.5开源NVIDIA	1,340	±10	3,625	—
25	—	gemma-3-12b-it开源Google	1,339	±10	3,866	—
26	—	qwq-32b开源Alibaba · 另有 1 个档位	1,334	±5	23,639	—
27	—	llama-3.1-405b-instruct-bf16开源Meta	1,333	±4	41,932	—
28	—	llama-3.1-405b-instruct-fp8开源Meta	1,331	±3	60,272	—
29	—	qwen3-30b-a3b开源Alibaba	1,326	±5	24,682	—
30	—	gpt-oss-20b开源OpenAI	1,325	±7	7,197	—
31	—	llama-4-maverick-17b-128e-instruct开源Meta	1,325	±4	36,588	—
32	—	llama-3.3-nemotron-49b-super-v1开源NVIDIA	1,324	±12	2,243	—
33	—	deepseek-v2.5-1210开源DeepSeek · 另有 1 个档位	1,320	±8	6,877	—
34	—	llama-4-scout-17b-16e-instruct开源Meta	1,319	±5	26,364	—
35	—	gemma-3n-e4b-it开源Google	1,318	±5	24,134	—
36	—	llama-3.3-70b-instruct开源Meta	1,316	±3	52,593	—
37	—	qwen-max-0919开源Alibaba	1,315	±6	16,598	—
38	—	athene-v2-chat开源NexusFlow	1,311	±4	24,880	—
39	—	mistral-large-2407开源Mistral · 另有 1 个档位	1,311	±4	45,968	—
40	—	athene-70b-0725开源NexusFlow	1,302	±6	19,796	—
41	—	gemma-3-4b-it开源Google	1,301	±9	4,195	—
42	—	qwen2.5-72b-instruct开源Alibaba	1,299	±4	39,632	—
43	—	mistral-small-3.1-24b-instruct-2503开源Mistral	1,296	±5	26,077	—
44	—	llama-3.1-nemotron-70b-instruct开源NVIDIA	1,296	±8	7,216	—
45	—	llama-3.1-70b-instruct开源Meta	1,291	±4	56,003	—
46	—	jamba-1.5-large开源AI21 Labs	1,285	±7	8,730	—
47	—	llama-3.1-tulu-3-70b开源Ai2	1,284	±10	2,881	—
48	—	gemma-2-27b-it开源Google	1,284	±3	76,195	—
49	—	llama-3.1-nemotron-51b-instruct开源NVIDIA	1,284	±10	3,777	—
50	—	gemma-2-9b-it-simpo开源Princeton	1,276	±7	10,108	—
51	—	nemotron-4-340b-instruct开源NVIDIA	1,276	±5	19,913	—
52	—	command-r-plus-08-2024开源Cohere	1,275	±6	9,931	—
53	—	llama-3-70b-instruct开源Meta	1,273	±3	158,908	—
54	—	mistral-small-24b-instruct-2501开源Mistral	1,271	±6	14,830	—
55	—	qwen2.5-coder-32b-instruct开源Alibaba	1,267	±8	5,452	—
56	—	c4ai-aya-expanse-32b开源Cohere	1,265	±5	27,362	—
57	—	gemma-2-9b-it开源Google	1,261	±4	54,954	—
58	—	command-r-plus开源Cohere	1,261	±4	78,401	—
59	—	deepseek-coder-v2开源DeepSeek	1,260	±6	15,242	—
60	—	qwen2-72b-instruct开源Alibaba	1,260	±5	37,688	—
61	—	phi-4开源Microsoft	1,253	±5	24,354	—
62	—	olmo-2-0325-32b-instruct开源Ai2	1,251	±11	3,377	—
63	—	command-r-08-2024开源Cohere	1,249	±7	10,229	—
64	—	jamba-1.5-mini开源AI21 Labs	1,236	±7	8,918	—
65	—	ministral-8b-2410开源Mistral	1,234	±9	4,833	—
66	—	qwen1.5-110b-chat开源Alibaba	1,232	±5	26,679	—
67	—	qwen1.5-72b-chat开源Alibaba	1,231	±5	39,689	—
68	—	mixtral-8x22b-instruct-v0.1开源Mistral	1,227	±4	52,214	—
69	—	command-r开源Cohere	1,226	±5	54,710	—
70	—	c4ai-aya-expanse-8b开源Cohere	1,221	±7	9,922	—
71	—	llama-3-8b-instruct开源Meta	1,221	±4	106,055	—
72	—	llama-3.1-tulu-3-8b开源Ai2	1,221	±11	2,943	—
73	—	yi-1.5-34b-chat开源01 AI	1,212	±5	24,417	—
74	—	zephyr-orpo-141b-A35b-v0.1开源HuggingFace	1,211	±11	4,712	—
75	—	llama-3.1-8b-instruct开源Meta	1,210	±4	50,234	—
76	—	granite-3.1-8b-instruct开源IBM	1,208	±11	3,142	—
77	—	qwen1.5-32b-chat开源Alibaba	1,203	±6	22,068	—
78	—	phi-3-medium-4k-instruct开源Microsoft	1,196	±5	25,301	—
79	—	mixtral-8x7b-instruct-v0.1开源Mistral	1,196	±4	74,303	—
80	—	gemma-2-2b-it开源Google	1,195	±4	46,901	—
81	—	dbrx-instruct-preview开源Databricks	1,193	±6	32,760	—
82	—	internlm2_5-20b-chat开源InternLM	1,191	±7	10,038	—
83	—	qwen1.5-14b-chat开源Alibaba	1,190	±7	18,066	—
84	—	wizardlm-70b开源Microsoft	1,183	±9	8,270	—
85	—	yi-34b-chat开源01 AI	1,182	±7	15,624	—
86	—	deepseek-llm-67b-chat开源DeepSeek	1,181	±11	4,950	—
87	—	granite-3.0-8b-instruct开源IBM	1,180	±9	6,727	—
88	—	openchat-3.5-0106开源OpenChat · 另有 1 个档位	1,180	±8	12,712	—
89	—	granite-3.1-2b-instruct开源IBM	1,179	±11	3,235	—
90	—	snowflake-arctic-instruct开源Snowflake	1,177	±6	33,272	—
91	—	tulu-2-dpo-70b开源AllenAI/UW	1,177	±10	6,579	—
92	—	gemma-1.1-7b-it开源Google	1,175	±6	24,327	—
93	—	openhermes-2.5-mistral-7b开源NousResearch	1,173	±10	5,026	—
94	—	starling-lm-7b-beta开源Nexusflow	1,171	±7	16,190	—
95	—	vicuna-33b开源LMSYS	1,171	±6	22,613	—
96	—	phi-3-small-8k-instruct开源Microsoft	1,170	±6	17,983	—
97	—	llama-2-70b-chat开源Meta	1,169	±5	38,767	—
98	—	starling-lm-7b-alpha开源UC Berkeley	1,166	±8	10,267	—
99	—	llama-3.2-3b-instruct开源Meta	1,165	±8	8,043	—
100	—	nous-hermes-2-mixtral-8x7b-dpo开源NousResearch	1,164	±12	3,792	—
101	—	llama2-70b-steerlm-chat开源NVIDIA	1,155	±13	3,605	—
102	—	granite-3.0-2b-instruct开源IBM	1,154	±8	6,922	—
103	—	solar-10.7b-instruct-v1.0开源Upstage AI	1,152	±13	4,187	—
104	—	mistral-7b-instruct-v0.2开源Mistral	1,149	±7	19,603	—
105	—	mpt-30b-chat开源MosaicML	1,148	±12	2,606	—
106	—	wizardlm-13b开源Microsoft	1,148	±9	7,122	—
107	—	qwen1.5-7b-chat开源Alibaba	1,142	±10	4,782	—
108	—	phi-3-mini-4k-instruct-june-2024开源Microsoft	1,141	±6	12,415	—
109	—	llama-2-13b-chat开源Meta	1,140	±7	19,357	—
110	—	vicuna-13b开源LMSYS	1,139	±7	19,539	—
111	—	qwen-14b-chat开源Alibaba	1,136	±11	5,004	—
112	—	codellama-34b-instruct开源Meta	1,134	±9	7,417	—
113	—	gemma-7b-it开源Google	1,131	±9	9,034	—
114	—	zephyr-7b-beta开源HuggingFace	1,130	±9	11,220	—
115	—	phi-3-mini-128k-instruct开源Microsoft	1,128	±7	21,024	—
116	—	phi-3-mini-4k-instruct开源Microsoft	1,127	±6	20,539	—
117	—	guanaco-33b开源UW	1,126	±12	2,955	—
118	—	stripedhyena-nous-7b开源Together AI	1,118	±11	5,214	—
119	—	smollm2-1.7b-instruct开源HuggingFace	1,117	±14	2,244	—
120	—	vicuna-7b开源LMSYS	1,112	±9	6,972	—
121	—	llama-3.2-1b-instruct开源Meta	1,111	±8	8,166	—
122	—	gemma-1.1-2b-it开源Google	1,109	±8	11,035	—
123	—	mistral-7b-instruct开源Mistral	1,107	±9	9,042	—
124	—	llama-2-7b-chat开源Meta	1,107	±7	14,272	—
125	—	qwen1.5-4b-chat开源Alibaba	1,088	±9	7,662	—
126	—	gemma-2b-it开源Google	1,087	±12	4,817	—
127	—	olmo-7b-instruct开源Ai2	1,073	±11	6,412	—
128	—	koala-13b开源UC Berkeley	1,068	±10	6,998	—
129	—	alpaca-13b开源Stanford	1,062	±11	5,828	—
130	—	mpt-7b-chat开源MosaicML	1,058	±12	3,977	—
131	—	chatglm3-6b开源Tsinghua	1,054	±12	4,692	—
132	—	RWKV-4-Raven-14B开源RWKV	1,039	±11	4,898	—
133	—	chatglm2-6b开源Tsinghua	1,022	±14	2,683	—
134	—	oasst-pythia-12b开源OpenAssistant	1,019	±11	6,343	—
135	—	chatglm-6b开源Tsinghua	994	±13	4,968	—
136	—	fastchat-t5-3b开源LMSYS	988	±12	4,270	—
137	—	dolly-v2-12b开源Databricks	975	±14	3,471	—
138	—	llama-13b开源Meta	966	±16	2,441	—
139	—	stablelm-tuned-alpha-7b开源Stability AI	951	±13	3,325	—

其余 89 个模型

智力进步

每家厂商当时最强的模型在 Artificial Analysis 智力指数上的分数，按模型发布的月份画。

涨得最多 · Anthropic +57.2目前最高 · Anthropic 64.6最高与最低相差 18.1 分时间范围 2025年11月 – 2026年11月

64.6Anthropic60.0Google56.9OpenAI50.1Meta46.5Kimi