模型排行

排名	變化	模型	得分	置信區間	投票	價格 $/M
1	—	glm-4.7開源Z.ai	1,443	±8	7,157	—
2	—	kimi-k2-thinking-turbo開源Moonshot AI	1,430	±5	23,702	—
3	—	glm-4.6開源Z.ai	1,424	±4	31,660	—
4	—	deepseek-v3.2-exp開源DeepSeek · 另有 1 個檔位	1,423	±6	11,910	—
5	—	qwen3-235b-a22b-instruct-2507開源Alibaba	1,422	±3	59,848	—
6	—	deepseek-v3.2開源DeepSeek · 另有 1 個檔位	1,420	±6	18,091	—
7	—	deepseek-r1-0528開源DeepSeek · 另有 1 個檔位	1,419	±6	19,149	—
8	—	kimi-k2-0905-preview開源Moonshot AI · 另有 1 個檔位	1,418	±6	12,061	—
9	—	deepseek-v3.1開源DeepSeek · 另有 1 個檔位	1,417	±6	15,189	—
10	—	deepseek-v3.1-terminus開源DeepSeek · 另有 1 個檔位	1,415	±10	3,731	—
11	▲ 1	qwen3-vl-235b-a22b-instruct開源Alibaba	1,414	±7	11,676	—
12	▼ 1	mistral-large-3開源Mistral	1,413	±6	14,472	—
13	—	glm-4.5開源Z.ai	1,409	±5	24,694	—
14	—	qwen3-next-80b-a3b-instruct開源Alibaba	1,401	±5	22,974	—
15	—	longcat-flash-chat開源Meituan	1,400	±6	11,455	—
16	—	qwen3-235b-a22b-no-thinking開源Alibaba	1,400	±4	39,205	—
17	—	qwen3-235b-a22b-thinking-2507開源Alibaba	1,398	±6	9,299	—
18	—	mimo-v2-flash (non-thinking)開源Xiaomi	1,397	±8	6,623	—
19	—	qwen3-vl-235b-a22b-thinking開源Alibaba	1,394	±7	7,951	—
20	—	deepseek-v3-0324開源DeepSeek · 另有 1 個檔位	1,393	±4	46,580	—
21	—	qwen3-coder-480b-a35b-instruct開源Alibaba	1,387	±5	26,656	—
22	▲ 1	minimax-m2.1-preview開源MiniMax	1,384	±8	6,148	—
23	▼ 1	qwen3-30b-a3b-instruct-2507開源Alibaba	1,382	±5	24,077	—
24	—	qwen3-235b-a22b開源Alibaba	1,374	±5	27,047	—
25	—	glm-4.5-air開源Z.ai	1,371	±4	31,472	—
26	—	qwen3-next-80b-a3b-thinking開源Alibaba	1,368	±6	13,750	—
27	—	minimax-m1開源MiniMax	1,367	±4	36,679	—
28	—	gemma-3-27b-it開源Google	1,365	±4	49,137	—
29	—	mistral-small-2506開源Mistral	1,355	±5	18,225	—
30	—	intellect-3開源Prime Intellect	1,354	±8	5,424	—
31	▲ 1	glm-4.5v開源Z.ai	1,353	±8	4,961	—
32	▼ 1	gpt-oss-120b開源OpenAI	1,353	±4	31,079	$0.26
33	—	command-a-03-2025開源Cohere	1,352	±3	57,583	—
34	—	ling-flash-2.0開源Ant Group	1,347	±7	7,111	—
35	▲ 1	minimax-m2開源MiniMax	1,346	±8	7,073	—
36	▼ 1	step-3開源StepFun	1,346	±7	6,600	—
37	—	llama-3.1-nemotron-ultra-253b-v1開源NVIDIA	1,346	±12	2,573	—
38	—	qwen3-32b開源Alibaba	1,345	±9	3,944	—
39	▲ 1	nvidia-llama-3.3-nemotron-super-49b-v1.5開源NVIDIA	1,341	±10	3,465	—
40	▼ 1	gemma-3-12b-it開源Google	1,341	±9	3,866	—
41	—	llama-3.1-405b-instruct-bf16開源Meta	1,335	±4	41,948	—
42	—	qwq-32b開源Alibaba · 另有 1 個檔位	1,334	±4	26,177	—
43	—	llama-3.1-405b-instruct-fp8開源Meta	1,333	±3	60,273	—
44	—	llama-4-maverick-17b-128e-instruct開源Meta	1,327	±4	41,039	—
45	—	qwen3-30b-a3b開源Alibaba	1,326	±5	27,383	—
46	—	llama-3.3-nemotron-49b-super-v1開源NVIDIA	1,326	±12	2,246	—
47	—	deepseek-v2.5-1210開源DeepSeek · 另有 1 個檔位	1,323	±8	6,877	—
48	—	llama-4-scout-17b-16e-instruct開源Meta	1,322	±5	31,044	—
49	—	ring-flash-2.0開源Ant Group	1,320	±7	7,243	—
50	—	llama-3.3-70b-instruct開源Meta	1,319	±3	55,942	—
51	—	gemma-3n-e4b-it開源Google	1,319	±5	23,401	—
52	—	gpt-oss-20b開源OpenAI	1,318	±6	10,820	—
53	▲ 1	qwen-max-0919開源Alibaba	1,317	±6	16,599	—
54	▼ 1	nvidia-nemotron-3-nano-30b-a3b-bf16開源NVIDIA	1,317	±7	9,342	—
55	—	mistral-large-2407開源Mistral · 另有 1 個檔位	1,314	±4	45,968	—
56	—	athene-v2-chat開源NexusFlow	1,313	±4	24,880	—
57	—	athene-70b-0725開源NexusFlow	1,305	±6	19,795	—
58	—	olmo-3-32b-think開源Ai2	1,305	±8	5,977	—
59	—	mistral-small-3.1-24b-instruct-2503開源Mistral	1,304	±4	33,996	—
60	—	gemma-3-4b-it開源Google	1,303	±9	4,196	—
61	—	qwen2.5-72b-instruct開源Alibaba	1,302	±4	39,634	—
62	—	llama-3.1-nemotron-70b-instruct開源NVIDIA	1,298	±8	7,216	—
63	—	llama-3.1-70b-instruct開源Meta	1,293	±4	56,002	—
64	—	jamba-1.5-large開源AI21 Labs	1,289	±7	8,730	—
65	—	ibm-granite-h-small開源IBM	1,288	±8	5,681	—
66	—	gemma-2-27b-it開源Google	1,288	±3	76,196	—
67	—	llama-3.1-nemotron-51b-instruct開源NVIDIA	1,286	±10	3,777	—
68	—	llama-3.1-tulu-3-70b開源Ai2	1,286	±10	2,882	—
69	—	gemma-2-9b-it-simpo開源Princeton	1,279	±7	10,108	—
70	—	nemotron-4-340b-instruct開源NVIDIA	1,277	±5	19,913	—
71	—	command-r-plus-08-2024開源Cohere	1,277	±7	9,932	—
72	—	llama-3-70b-instruct開源Meta	1,276	±4	158,907	—
73	—	mistral-small-24b-instruct-2501開源Mistral	1,274	±6	14,831	—
74	—	qwen2.5-coder-32b-instruct開源Alibaba	1,269	±8	5,452	—
75	—	c4ai-aya-expanse-32b開源Cohere	1,267	±5	27,362	—
76	—	gemma-2-9b-it開源Google	1,265	±4	54,957	—
77	—	deepseek-coder-v2開源DeepSeek	1,264	±6	15,242	—
78	—	command-r-plus開源Cohere	1,263	±4	78,400	—
79	—	qwen2-72b-instruct開源Alibaba	1,262	±5	37,688	—
80	—	phi-4開源Microsoft	1,256	±5	24,355	—
81	—	olmo-2-0325-32b-instruct開源Ai2	1,252	±11	3,379	—
82	—	command-r-08-2024開源Cohere	1,251	±7	10,229	—
83	—	jamba-1.5-mini開源AI21 Labs	1,239	±7	8,918	—
84	—	ministral-8b-2410開源Mistral	1,237	±9	4,833	—
85	—	qwen1.5-110b-chat開源Alibaba	1,234	±5	26,679	—
86	—	qwen1.5-72b-chat開源Alibaba	1,234	±5	39,689	—
87	—	mixtral-8x22b-instruct-v0.1開源Mistral	1,230	±4	52,214	—
88	—	command-r開源Cohere	1,228	±5	54,710	—
89	—	c4ai-aya-expanse-8b開源Cohere	1,224	±7	9,922	—
90	—	llama-3-8b-instruct開源Meta	1,223	±4	106,055	—
91	—	llama-3.1-tulu-3-8b開源Ai2	1,222	±11	2,943	—
92	—	zephyr-orpo-141b-A35b-v0.1開源HuggingFace	1,214	±11	4,712	—
93	—	yi-1.5-34b-chat開源01 AI	1,213	±5	24,417	—
94	—	llama-3.1-8b-instruct開源Meta	1,211	±4	50,235	—
95	—	granite-3.1-8b-instruct開源IBM	1,210	±11	3,142	—
96	—	qwen1.5-32b-chat開源Alibaba	1,205	±6	22,068	—
97	—	gemma-2-2b-it開源Google	1,199	±4	46,900	—
98	—	phi-3-medium-4k-instruct開源Microsoft	1,198	±5	25,301	—
99	—	mixtral-8x7b-instruct-v0.1開源Mistral	1,198	±4	74,303	—
100	—	dbrx-instruct-preview開源Databricks	1,196	±6	32,760	—
101	—	qwen1.5-14b-chat開源Alibaba	1,193	±7	18,066	—
102	—	internlm2_5-20b-chat開源InternLM	1,192	±7	10,038	—
103	—	wizardlm-70b開源Microsoft	1,185	±9	8,270	—
104	—	deepseek-llm-67b-chat開源DeepSeek	1,184	±12	4,950	—
105	—	yi-34b-chat開源01 AI	1,184	±7	15,624	—
106	—	granite-3.0-8b-instruct開源IBM	1,184	±9	6,727	—
107	—	openchat-3.5-0106開源OpenChat · 另有 1 個檔位	1,183	±8	12,712	—
108	—	granite-3.1-2b-instruct開源IBM	1,181	±11	3,236	—
109	▲ 1	gemma-1.1-7b-it開源Google	1,180	±6	24,327	—
110	▼ 1	snowflake-arctic-instruct開源Snowflake	1,180	±6	33,272	—
111	—	tulu-2-dpo-70b開源AllenAI/UW	1,179	±10	6,579	—
112	—	openhermes-2.5-mistral-7b開源NousResearch	1,176	±10	5,026	—
113	—	vicuna-33b開源LMSYS	1,174	±6	22,613	—
114	—	starling-lm-7b-beta開源Nexusflow	1,173	±7	16,190	—
115	—	phi-3-small-8k-instruct開源Microsoft	1,172	±6	17,983	—
116	—	llama-2-70b-chat開源Meta	1,172	±5	38,767	—
117	—	starling-lm-7b-alpha開源UC Berkeley	1,168	±8	10,267	—
118	—	llama-3.2-3b-instruct開源Meta	1,167	±8	8,043	—
119	—	nous-hermes-2-mixtral-8x7b-dpo開源NousResearch	1,166	±12	3,792	—
120	—	llama2-70b-steerlm-chat開源NVIDIA	1,157	±13	3,605	—
121	—	granite-3.0-2b-instruct開源IBM	1,157	±8	6,922	—
122	—	solar-10.7b-instruct-v1.0開源Upstage AI	1,154	±13	4,187	—
123	—	dolphin-2.2.1-mistral-7b開源Cognitive Computations	1,152	±15	1,685	—
124	—	mpt-30b-chat開源MosaicML	1,151	±12	2,606	—
125	—	mistral-7b-instruct-v0.2開源Mistral	1,151	±7	19,603	—
126	—	wizardlm-13b開源Microsoft	1,150	±9	7,122	—
127	—	falcon-180b-chat開源TII	1,148	±17	1,312	—
128	—	qwen1.5-7b-chat開源Alibaba	1,144	±10	4,782	—
129	—	phi-3-mini-4k-instruct-june-2024開源Microsoft	1,143	±6	12,414	—
130	—	llama-2-13b-chat開源Meta	1,143	±7	19,357	—
131	—	vicuna-13b開源LMSYS	1,142	±7	19,539	—
132	—	qwen-14b-chat開源Alibaba	1,139	±11	5,004	—
133	—	codellama-34b-instruct開源Meta	1,137	±9	7,417	—
134	—	gemma-7b-it開源Google	1,136	±9	9,034	—
135	—	zephyr-7b-beta開源HuggingFace	1,132	±9	11,220	—
136	—	phi-3-mini-128k-instruct開源Microsoft	1,131	±7	21,024	—
137	—	phi-3-mini-4k-instruct開源Microsoft	1,129	±6	20,539	—
138	—	zephyr-7b-alpha開源HuggingFace	1,129	±16	1,803	—
139	—	guanaco-33b開源UW	1,128	±12	2,955	—
140	—	stripedhyena-nous-7b開源Together AI	1,121	±11	5,214	—
141	—	codellama-70b-instruct開源Meta	1,119	±18	1,151	—
142	—	smollm2-1.7b-instruct開源HuggingFace	1,119	±14	2,244	—
143	—	vicuna-7b開源LMSYS	1,116	±9	6,972	—
144	—	gemma-1.1-2b-it開源Google	1,115	±8	11,035	—
145	—	llama-3.2-1b-instruct開源Meta	1,113	±8	8,166	—
146	—	mistral-7b-instruct開源Mistral	1,111	±9	9,042	—
147	—	llama-2-7b-chat開源Meta	1,109	±7	14,272	—
148	—	gemma-2b-it開源Google	1,092	±12	4,817	—
149	—	qwen1.5-4b-chat開源Alibaba	1,091	±9	7,662	—
150	—	olmo-7b-instruct開源Ai2	1,074	±11	6,412	—
151	—	koala-13b開源UC Berkeley	1,071	±10	6,998	—
152	—	alpaca-13b開源Stanford	1,067	±11	5,828	—
153	—	gpt4all-13b-snoozy開源Nomic AI	1,066	±15	1,773	—
154	—	mpt-7b-chat開源MosaicML	1,062	±12	3,977	—
155	—	chatglm3-6b開源Tsinghua	1,057	±12	4,692	—
156	—	RWKV-4-Raven-14B開源RWKV	1,042	±11	4,898	—
157	—	chatglm2-6b開源Tsinghua	1,026	±14	2,683	—
158	—	oasst-pythia-12b開源OpenAssistant	1,023	±11	6,343	—
159	—	chatglm-6b開源Tsinghua	996	±13	4,968	—
160	—	fastchat-t5-3b開源LMSYS	992	±12	4,270	—
161	—	dolly-v2-12b開源Databricks	980	±14	3,471	—
162	—	llama-13b開源Meta	972	±16	2,441	—
163	—	stablelm-tuned-alpha-7b開源Stability AI	953	±13	3,325	—

其餘 113 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