模型排行

排名	變化	模型	得分	置信區間	投票	價格 $/M
1	—	deepseek-r1開源DeepSeek	1,316	±10	3,286	—
2	—	deepseek-v3開源DeepSeek	1,275	±6	14,628	—
3	—	llama-3.1-405b-instruct-bf16開源Meta	1,254	±5	23,439	—
4	—	llama-3.1-405b-instruct-fp8開源Meta	1,253	±3	63,177	—
5	—	deepseek-v2.5-1210開源DeepSeek · 另有 1 個檔位	1,240	±6	7,261	—
6	—	llama-3.3-70b-instruct開源Meta	1,235	±4	17,718	—
7	—	athene-v2-chat開源NexusFlow	1,232	±4	23,145	—
8	—	mistral-large-2407開源Mistral · 另有 1 個檔位	1,232	±3	48,196	—
9	—	qwen-max-0919開源Alibaba	1,232	±4	17,435	—
10	—	athene-70b-0725開源NexusFlow	1,223	±4	20,608	—
11	—	qwen2.5-72b-instruct開源Alibaba	1,222	±3	41,073	—
12	—	llama-3.1-nemotron-70b-instruct開源NVIDIA	1,217	±7	7,597	—
13	—	llama-3.1-70b-instruct開源Meta	1,214	±3	58,787	—
14	—	llama-3.1-tulu-3-70b開源Ai2	1,209	±10	3,024	—
15	—	gemma-2-27b-it開源Google	1,208	±3	73,585	—
16	—	jamba-1.5-large開源AI21 Labs	1,205	±6	9,128	—
17	—	llama-3.1-nemotron-51b-instruct開源NVIDIA	1,204	±9	3,898	—
18	—	gemma-2-9b-it-simpo開源Princeton	1,200	±5	10,553	—
19	—	command-r-plus-08-2024開源Cohere	1,198	±5	10,536	—
20	—	nemotron-4-340b-instruct開源NVIDIA	1,196	±4	20,604	—
21	—	llama-3-70b-instruct開源Meta	1,195	±2	163,796	—
22	—	qwen2.5-coder-32b-instruct開源Alibaba	1,190	±8	5,731	—
23	—	c4ai-aya-expanse-32b開源Cohere	1,185	±4	28,157	—
24	▲ 1	command-r-plus開源Cohere	1,183	±3	80,877	—
25	▲ 1	deepseek-coder-v2開源DeepSeek	1,182	±4	15,755	—
26	▼ 2	phi-4開源Microsoft	1,181	±8	4,965	—
27	—	qwen2-72b-instruct開源Alibaba	1,181	±3	38,873	—
28	—	gemma-2-9b-it開源Google	1,180	±3	51,362	—
29	—	command-r-08-2024開源Cohere	1,176	±5	10,848	—
30	—	jamba-1.5-mini開源AI21 Labs	1,160	±7	9,270	—
31	—	ministral-8b-2410開源Mistral	1,160	±7	5,116	—
32	—	qwen1.5-110b-chat開源Alibaba	1,156	±4	27,464	—
33	—	c4ai-aya-expanse-8b開源Cohere	1,151	±6	9,991	—
34	—	mixtral-8x22b-instruct-v0.1開源Mistral	1,149	±3	53,782	—
35	—	qwen1.5-72b-chat開源Alibaba	1,148	±3	40,638	—
36	—	command-r開源Cohere	1,148	±2	56,375	—
37	—	llama-3.1-tulu-3-8b開源Ai2	1,147	±9	3,076	—
38	—	llama-3-8b-instruct開源Meta	1,143	±3	109,221	—
39	—	granite-3.1-8b-instruct開源IBM	1,135	±10	3,305	—
40	—	yi-1.5-34b-chat開源01 AI	1,134	±4	25,126	—
41	—	llama-3.1-8b-instruct開源Meta	1,133	±3	52,658	—
42	—	zephyr-orpo-141b-A35b-v0.1開源HuggingFace	1,131	±7	4,862	—
43	—	qwen1.5-32b-chat開源Alibaba	1,127	±4	22,762	—
44	—	phi-3-medium-4k-instruct開源Microsoft	1,118	±4	26,116	—
45	—	gemma-2-2b-it開源Google	1,117	±4	43,006	—
46	—	internlm2_5-20b-chat開源InternLM	1,116	±6	10,594	—
47	—	mixtral-8x7b-instruct-v0.1開源Mistral	1,114	±0	76,130	—
48	—	dbrx-instruct-preview開源Databricks	1,113	±4	33,734	—
49	—	granite-3.0-8b-instruct開源IBM	1,109	±7	7,000	—
50	—	qwen1.5-14b-chat開源Alibaba	1,107	±4	18,672	—
51	—	wizardlm-70b開源Microsoft	1,105	±8	8,380	—
52	—	granite-3.1-2b-instruct開源IBM	1,103	±11	3,389	—
53	—	snowflake-arctic-instruct開源Snowflake	1,102	±4	34,168	—
54	—	yi-34b-chat開源01 AI	1,101	±5	15,923	—
55	—	openchat-3.5-0106開源OpenChat · 另有 1 個檔位	1,101	±5	12,988	—
56	—	gemma-1.1-7b-it開源Google	1,099	±4	25,062	—
57	—	phi-3-small-8k-instruct開源Microsoft	1,098	±4	18,477	—
58	—	llama-3.2-3b-instruct開源Meta	1,094	±6	8,403	—
59	—	tulu-2-dpo-70b開源AllenAI/UW	1,092	±6	6,663	—
60	—	starling-lm-7b-beta開源Nexusflow	1,092	±4	16,673	—
61	—	vicuna-33b開源LMSYS	1,092	±3	22,949	—
62	—	deepseek-llm-67b-chat開源DeepSeek	1,091	±8	4,987	—
63	—	openhermes-2.5-mistral-7b開源NousResearch	1,090	±8	5,091	—
64	—	qwq-32b-preview開源Alibaba	1,088	±10	3,414	—
65	—	starling-lm-7b-alpha開源UC Berkeley	1,083	±6	10,416	—
66	—	granite-3.0-2b-instruct開源IBM	1,082	±8	7,193	—
67	—	llama-2-70b-chat開源Meta	1,082	±3	39,628	—
68	—	llama2-70b-steerlm-chat開源NVIDIA	1,074	±8	3,637	—
69	—	nous-hermes-2-mixtral-8x7b-dpo開源NousResearch	1,072	±8	3,834	—
70	—	dolphin-2.2.1-mistral-7b開源Cognitive Computations	1,068	±12	1,713	—
71	—	phi-3-mini-4k-instruct-june-2024開源Microsoft	1,068	±6	12,817	—
72	—	mistral-7b-instruct-v0.2開源Mistral	1,067	±4	20,054	—
73	—	solar-10.7b-instruct-v1.0開源Upstage AI	1,066	±9	4,288	—
74	—	qwen1.5-7b-chat開源Alibaba	1,065	±8	4,869	—
75	—	wizardlm-13b開源Microsoft	1,063	±8	7,177	—
76	▲ 1	mpt-30b-chat開源MosaicML	1,060	±10	2,647	—
77	▼ 1	falcon-180b-chat開源TII	1,059	±16	1,326	—
78	—	vicuna-13b開源LMSYS	1,058	±6	19,790	—
79	—	qwen-14b-chat開源Alibaba	1,057	±7	5,073	—
80	—	phi-3-mini-4k-instruct開源Microsoft	1,056	±4	21,094	—
81	—	llama-2-13b-chat開源Meta	1,054	±5	19,736	—
82	—	smollm2-1.7b-instruct開源HuggingFace	1,052	±13	2,372	—
83	—	zephyr-7b-beta開源HuggingFace	1,049	±5	11,333	—
84	—	phi-3-mini-128k-instruct開源Microsoft	1,049	±4	21,620	—
85	—	gemma-7b-it開源Google	1,048	±7	9,176	—
86	—	codellama-34b-instruct開源Meta	1,047	±8	7,511	—
87	—	zephyr-7b-alpha開源HuggingFace	1,046	±12	1,814	—
88	—	llama-3.2-1b-instruct開源Meta	1,045	±6	8,532	—
89	—	gemma-1.1-2b-it開源Google	1,040	±5	11,340	—
90	▲ 1	guanaco-33b開源UW	1,040	±12	2,999	—
91	▼ 1	codellama-70b-instruct開源Meta	1,037	±17	1,193	—
92	—	stripedhyena-nous-7b開源Together AI	1,034	±8	5,272	—
93	—	mistral-7b-instruct開源Mistral	1,027	±7	9,143	—
94	—	llama-2-7b-chat開源Meta	1,025	±5	14,551	—
95	—	vicuna-7b開源LMSYS	1,021	±7	7,014	—
96	—	qwen1.5-4b-chat開源Alibaba	1,015	±7	7,814	—
97	—	gemma-2b-it開源Google	1,009	±9	4,923	—
98	—	olmo-7b-instruct開源Ai2	994	±7	6,508	—
99	—	koala-13b開源UC Berkeley	979	±8	7,036	—
100	—	chatglm3-6b開源Tsinghua	974	±8	4,763	—
101	—	gpt4all-13b-snoozy開源Nomic AI	972	±14	1,786	—
102	—	mpt-7b-chat開源MosaicML	967	±8	4,013	—
103	—	alpaca-13b開源Stanford	966	±7	5,877	—
104	—	RWKV-4-Raven-14B開源RWKV	949	±9	4,933	—
105	—	chatglm2-6b開源Tsinghua	942	±10	2,709	—
106	—	oasst-pythia-12b開源OpenAssistant	931	±9	6,380	—
107	—	fastchat-t5-3b開源LMSYS	897	±10	4,299	—
108	—	chatglm-6b開源Tsinghua	888	±9	4,988	—
109	—	dolly-v2-12b開源Databricks	881	±10	3,485	—
110	—	llama-13b開源Meta	867	±13	2,444	—
111	—	stablelm-tuned-alpha-7b開源Stability AI	859	±9	3,341	—

其餘 61 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