模型排行

排名	變化	模型	得分	置信區間	投票	價格 $/M
1	—	deepseek-v3-0324開源DeepSeek · 另有 1 個檔位	1,323	±6	8,753	—
2	—	deepseek-r1開源DeepSeek	1,316	±4	18,493	—
3	—	qwen3-235b-a22b開源Alibaba	1,286	±9	3,611	—
4	—	gemma-3-27b-it開源Google	1,281	±5	12,343	—
5	—	command-a-03-2025開源Cohere	1,265	±5	10,513	—
6	—	gemma-3-12b-it開源Google	1,263	±9	3,016	—
7	—	qwq-32b開源Alibaba · 另有 1 個檔位	1,256	±5	9,946	—
8	—	llama-3.1-405b-instruct-bf16開源Meta	1,254	±3	43,799	—
9	—	llama-3.1-405b-instruct-fp8開源Meta	1,253	±2	63,030	—
10	—	llama-4-maverick-17b-128e-instruct開源Meta	1,249	±6	7,798	—
11	—	llama-3.3-nemotron-49b-super-v1開源NVIDIA	1,247	±10	2,366	—
12	—	deepseek-v2.5-1210開源DeepSeek · 另有 1 個檔位	1,241	±5	7,242	—
13	—	llama-3.3-70b-instruct開源Meta	1,235	±3	38,088	—
14	—	mistral-large-2407開源Mistral · 另有 1 個檔位	1,232	±2	48,217	—
15	—	qwen-max-0919開源Alibaba	1,232	±4	17,432	—
16	—	athene-v2-chat開源NexusFlow	1,231	±3	26,077	—
17	—	athene-70b-0725開源NexusFlow	1,223	±4	20,578	—
18	—	qwen2.5-72b-instruct開源Alibaba	1,223	±2	41,517	—
19	—	gemma-3-4b-it開源Google	1,222	±10	3,326	—
20	—	llama-3.1-nemotron-70b-instruct開源NVIDIA	1,217	±6	7,579	—
21	—	llama-3.1-70b-instruct開源Meta	1,214	±2	58,637	—
22	—	llama-3.1-tulu-3-70b開源Ai2	1,209	±9	3,009	—
23	—	gemma-2-27b-it開源Google	1,207	±2	79,527	—
24	—	jamba-1.5-large開源AI21 Labs	1,206	±5	9,125	—
25	—	llama-3.1-nemotron-51b-instruct開源NVIDIA	1,204	±8	3,890	—
26	—	gemma-2-9b-it-simpo開源Princeton	1,200	±4	10,549	—
27	—	command-r-plus-08-2024開源Cohere	1,197	±5	10,537	—
28	—	nemotron-4-340b-instruct開源NVIDIA	1,196	±4	20,610	—
29	—	llama-3-70b-instruct開源Meta	1,195	±2	163,642	—
30	—	mistral-small-24b-instruct-2501開源Mistral	1,192	±4	15,323	—
31	—	qwen2.5-coder-32b-instruct開源Alibaba	1,190	±7	5,733	—
32	—	c4ai-aya-expanse-32b開源Cohere	1,185	±3	28,757	—
33	—	command-r-plus開源Cohere	1,183	±2	80,856	—
34	—	deepseek-coder-v2開源DeepSeek	1,182	±4	15,756	—
35	—	gemma-2-9b-it開源Google	1,181	±3	57,201	—
36	—	qwen2-72b-instruct開源Alibaba	1,181	±3	38,870	—
37	—	phi-4開源Microsoft	1,178	±3	25,221	—
38	—	command-r-08-2024開源Cohere	1,176	±5	10,849	—
39	—	olmo-2-0325-32b-instruct開源Ai2	1,176	±10	3,183	—
40	—	jamba-1.5-mini開源AI21 Labs	1,160	±5	9,273	—
41	—	ministral-8b-2410開源Mistral	1,160	±7	5,109	—
42	—	qwen1.5-110b-chat開源Alibaba	1,155	±4	27,438	—
43	—	c4ai-aya-expanse-8b開源Cohere	1,151	±6	10,395	—
44	—	mixtral-8x22b-instruct-v0.1開源Mistral	1,149	±2	53,761	—
45	—	qwen1.5-72b-chat開源Alibaba	1,148	±3	40,665	—
46	—	command-r開源Cohere	1,148	±3	56,392	—
47	—	llama-3.1-tulu-3-8b開源Ai2	1,147	±10	3,075	—
48	—	llama-3-8b-instruct開源Meta	1,143	±2	109,093	—
49	—	granite-3.1-8b-instruct開源IBM	1,135	±9	3,290	—
50	—	yi-1.5-34b-chat開源01 AI	1,134	±3	25,136	—
51	—	llama-3.1-8b-instruct開源Meta	1,133	±3	52,576	—
52	—	zephyr-orpo-141b-A35b-v0.1開源HuggingFace	1,131	±8	4,853	—
53	—	qwen1.5-32b-chat開源Alibaba	1,126	±4	22,762	—
54	—	phi-3-medium-4k-instruct開源Microsoft	1,118	±4	26,112	—
55	—	gemma-2-2b-it開源Google	1,118	±3	48,901	—
56	—	internlm2_5-20b-chat開源InternLM	1,116	±5	10,596	—
57	—	mixtral-8x7b-instruct-v0.1開源Mistral	1,114	±2	76,134	—
58	—	dbrx-instruct-preview開源Databricks	1,113	±3	33,743	—
59	—	granite-3.0-8b-instruct開源IBM	1,110	±6	7,002	—
60	—	qwen1.5-14b-chat開源Alibaba	1,107	±4	18,685	—
61	—	wizardlm-70b開源Microsoft	1,105	±7	8,383	—
62	—	granite-3.1-2b-instruct開源IBM	1,104	±9	3,380	—
63	—	snowflake-arctic-instruct開源Snowflake	1,102	±4	34,181	—
64	—	yi-34b-chat開源01 AI	1,101	±4	15,918	—
65	—	openchat-3.5-0106開源OpenChat · 另有 1 個檔位	1,100	±5	12,986	—
66	—	gemma-1.1-7b-it開源Google	1,099	±3	25,063	—
67	—	phi-3-small-8k-instruct開源Microsoft	1,098	±4	18,477	—
68	—	llama-3.2-3b-instruct開源Meta	1,093	±6	8,394	—
69	—	starling-lm-7b-beta開源Nexusflow	1,092	±5	16,674	—
70	—	tulu-2-dpo-70b開源AllenAI/UW	1,092	±7	6,658	—
71	—	vicuna-33b開源LMSYS	1,092	±4	22,937	—
72	—	deepseek-llm-67b-chat開源DeepSeek	1,091	±8	4,988	—
73	—	openhermes-2.5-mistral-7b開源NousResearch	1,089	±8	5,086	—
74	—	starling-lm-7b-alpha開源UC Berkeley	1,083	±6	10,414	—
75	—	granite-3.0-2b-instruct開源IBM	1,082	±7	7,187	—
76	—	llama-2-70b-chat開源Meta	1,082	±3	39,602	—
77	—	llama2-70b-steerlm-chat開源NVIDIA	1,074	±10	3,636	—
78	—	nous-hermes-2-mixtral-8x7b-dpo開源NousResearch	1,072	±9	3,837	—
79	—	dolphin-2.2.1-mistral-7b開源Cognitive Computations	1,069	±13	1,713	—
80	—	phi-3-mini-4k-instruct-june-2024開源Microsoft	1,068	±5	12,800	—
81	—	mistral-7b-instruct-v0.2開源Mistral	1,067	±4	20,063	—
82	—	solar-10.7b-instruct-v1.0開源Upstage AI	1,066	±10	4,286	—
83	—	qwen1.5-7b-chat開源Alibaba	1,065	±8	4,873	—
84	—	wizardlm-13b開源Microsoft	1,064	±7	7,175	—
85	—	mpt-30b-chat開源MosaicML	1,060	±11	2,644	—
86	—	falcon-180b-chat開源TII	1,059	±15	1,327	—
87	—	vicuna-13b開源LMSYS	1,058	±5	19,773	—
88	—	phi-3-mini-4k-instruct開源Microsoft	1,056	±4	21,092	—
89	—	qwen-14b-chat開源Alibaba	1,056	±9	5,067	—
90	—	llama-2-13b-chat開源Meta	1,054	±4	19,713	—
91	—	smollm2-1.7b-instruct開源HuggingFace	1,051	±12	2,376	—
92	—	zephyr-7b-beta開源HuggingFace	1,049	±5	11,321	—
93	—	phi-3-mini-128k-instruct開源Microsoft	1,049	±4	21,625	—
94	—	gemma-7b-it開源Google	1,048	±6	9,176	—
95	—	codellama-34b-instruct開源Meta	1,047	±6	7,508	—
96	—	zephyr-7b-alpha開源HuggingFace	1,046	±12	1,812	—
97	—	llama-3.2-1b-instruct開源Meta	1,045	±6	8,518	—
98	—	gemma-1.1-2b-it開源Google	1,040	±5	11,349	—
99	—	guanaco-33b開源UW	1,040	±11	2,995	—
100	—	codellama-70b-instruct開源Meta	1,038	±16	1,192	—
101	—	stripedhyena-nous-7b開源Together AI	1,034	±8	5,277	—
102	—	mistral-7b-instruct開源Mistral	1,026	±5	9,144	—
103	—	llama-2-7b-chat開源Meta	1,025	±4	14,532	—
104	—	vicuna-7b開源LMSYS	1,021	±8	7,015	—
105	—	qwen1.5-4b-chat開源Alibaba	1,014	±7	7,819	—
106	—	gemma-2b-it開源Google	1,008	±8	4,918	—
107	—	olmo-7b-instruct開源Ai2	994	±7	6,501	—
108	—	koala-13b開源UC Berkeley	980	±8	7,020	—
109	—	chatglm3-6b開源Tsinghua	974	±9	4,762	—
110	—	gpt4all-13b-snoozy開源Nomic AI	973	±14	1,788	—
111	—	mpt-7b-chat開源MosaicML	967	±10	3,997	—
112	—	alpaca-13b開源Stanford	967	±8	5,864	—
113	—	RWKV-4-Raven-14B開源RWKV	949	±8	4,920	—
114	—	chatglm2-6b開源Tsinghua	942	±10	2,712	—
115	—	oasst-pythia-12b開源OpenAssistant	931	±8	6,368	—
116	—	fastchat-t5-3b開源LMSYS	897	±9	4,288	—
117	—	chatglm-6b開源Tsinghua	889	±9	4,983	—
118	—	dolly-v2-12b開源Databricks	882	±11	3,480	—
119	—	llama-13b開源Meta	868	±15	2,446	—
120	—	stablelm-tuned-alpha-7b開源Stability AI	859	±10	3,336	—

其餘 70 個模型

智力進步

每家廠商當時最強的模型在 Artificial Analysis 智力指數上的分數，按模型釋出的月份畫。

漲得最多 · Anthropic +57.2目前最高 · Anthropic 64.6最高與最低相差 18.1 分時間範圍 2025年11月 – 2026年11月

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