模型排行

排名	變化	模型	得分	置信區間	投票	價格 $/M
1	—	glm-5開源Z.ai	1,455	±6	11,093	—
2	—	kimi-k2.5-thinking開源Moonshot AI	1,453	±5	16,262	—
3	—	qwen3.5-397b-a17b開源Alibaba	1,452	±6	10,431	—
4	—	glm-4.7開源Z.ai	1,443	±6	12,242	—
5	—	kimi-k2.5-instant開源Moonshot AI	1,433	±7	8,257	—
6	—	kimi-k2-thinking-turbo開源Moonshot AI	1,430	±4	41,738	—
7	—	glm-4.6開源Z.ai	1,426	±4	36,102	—
8	—	deepseek-v3.2-exp-thinking開源DeepSeek · 另有 1 個檔位	1,425	±7	9,188	—
9	—	deepseek-v3.2開源DeepSeek · 另有 1 個檔位	1,425	±4	36,511	—
10	—	qwen3-235b-a22b-instruct-2507開源Alibaba	1,422	±3	77,683	—
11	—	deepseek-r1-0528開源DeepSeek · 另有 1 個檔位	1,421	±6	18,831	—
12	▲ 1	deepseek-v3.1開源DeepSeek · 另有 1 個檔位	1,418	±6	15,150	—
13	▲ 1	kimi-k2-0905-preview開源Moonshot AI · 另有 1 個檔位	1,418	±6	11,924	—
14	▼ 2	qwen3.5-122b-a10b開源Alibaba	1,417	±7	6,946	—
15	▲ 1	deepseek-v3.1-terminus-thinking開源DeepSeek · 另有 1 個檔位	1,416	±10	3,497	—
16	▼ 1	mistral-large-3開源Mistral	1,416	±4	33,200	—
17	—	qwen3-vl-235b-a22b-instruct開源Alibaba	1,415	±6	11,645	—
18	—	glm-4.5開源Z.ai	1,411	±5	24,640	—
19	—	qwen3.5-27b開源Alibaba	1,406	±7	6,957	—
20	—	minimax-m2.5開源MiniMax	1,405	±6	11,909	—
21	—	qwen3-235b-a22b-no-thinking開源Alibaba	1,403	±4	38,797	—
22	—	qwen3-next-80b-a3b-instruct開源Alibaba	1,401	±5	23,187	—
23	▲ 2	qwen3.5-35b-a3b開源Alibaba	1,401	±7	7,278	—
24	▼ 1	longcat-flash-chat開源Meituan	1,400	±6	11,517	—
25	▼ 1	qwen3-235b-a22b-thinking-2507開源Alibaba	1,399	±6	9,128	—
26	—	qwen3-vl-235b-a22b-thinking開源Alibaba	1,396	±7	8,052	—
27	—	deepseek-v3-0324開源DeepSeek · 另有 1 個檔位	1,394	±4	46,144	—
28	—	mimo-v2-flash (non-thinking)開源Xiaomi · 另有 1 個檔位	1,392	±4	25,427	—
29	—	step-3.5-flash開源StepFun	1,389	±6	13,885	—
30	—	qwen3-coder-480b-a35b-instruct開源Alibaba	1,387	±5	26,162	—
31	—	minimax-m2.1-preview開源MiniMax	1,386	±5	17,271	—
32	—	qwen3-30b-a3b-instruct-2507開源Alibaba	1,383	±5	24,086	—
33	—	glm-4.6v開源Z.ai	1,378	±11	2,836	—
34	▲ 1	trinity-large開源Arcee AI	1,376	±7	8,033	—
35	▼ 1	qwen3-235b-a22b開源Alibaba	1,375	±5	26,679	—
36	—	glm-4.5-air開源Z.ai	1,372	±4	31,546	—
37	—	qwen3-next-80b-a3b-thinking開源Alibaba	1,369	±6	13,906	—
38	—	glm-4.7-flash開源Z.ai	1,368	±6	11,857	—
39	▲ 1	gemma-3-27b-it開源Google	1,365	±4	48,195	—
40	▲ 1	nvidia-nemotron-3-super-120b-a12b開源NVIDIA	1,365	±10	3,641	—
41	▼ 2	minimax-m1開源MiniMax	1,364	±4	35,951	—
42	▲ 1	mistral-small-2506開源Mistral	1,357	±5	18,015	—
43	▼ 1	intellect-3開源Prime Intellect	1,357	±8	5,387	—
44	—	gpt-oss-120b開源OpenAI	1,354	±4	31,077	$0.26
45	—	command-a-03-2025開源Cohere	1,354	±3	57,068	—
46	—	glm-4.5v開源Z.ai	1,353	±8	4,989	—
47	—	step-3開源StepFun	1,348	±7	6,600	—
48	▲ 3	minimax-m2開源MiniMax	1,347	±8	6,950	—
49	▲ 1	qwen3-32b開源Alibaba	1,347	±9	3,926	—
50	▼ 2	llama-3.1-nemotron-ultra-253b-v1開源NVIDIA	1,347	±12	2,549	—
51	▼ 2	ling-flash-2.0開源Ant Group	1,346	±7	7,125	—
52	—	nvidia-llama-3.3-nemotron-super-49b-v1.5開源NVIDIA	1,343	±10	3,392	—
53	—	gemma-3-12b-it開源Google	1,342	±10	3,829	—
54	—	qwq-32b開源Alibaba · 另有 1 個檔位	1,336	±4	25,749	—
55	—	llama-3.1-405b-instruct-bf16開源Meta	1,335	±4	41,375	—
56	—	llama-3.1-405b-instruct-fp8開源Meta	1,333	±4	59,656	—
57	—	olmo-3.1-32b-instruct開源Ai2	1,331	±6	12,326	—
58	▲ 1	qwen3-30b-a3b開源Alibaba	1,328	±5	26,908	—
59	▲ 2	llama-3.3-nemotron-49b-super-v1開源NVIDIA	1,327	±12	2,218	—
60	—	llama-4-maverick-17b-128e-instruct開源Meta	1,327	±4	40,529	—
61	▼ 3	molmo-2-8b開源Ai2	1,326	±21	811	—
62	—	deepseek-v2.5-1210開源DeepSeek · 另有 1 個檔位	1,323	±8	6,795	—
63	—	llama-4-scout-17b-16e-instruct開源Meta	1,322	±5	30,740	—
64	—	ring-flash-2.0開源Ant Group	1,321	±7	7,252	—
65	—	gemma-3n-e4b-it開源Google	1,318	±5	22,915	—
66	—	llama-3.3-70b-instruct開源Meta	1,318	±3	55,149	—
67	▲ 2	gpt-oss-20b開源OpenAI	1,318	±6	10,750	—
68	—	nvidia-nemotron-3-nano-30b-a3b-bf16開源NVIDIA	1,318	±6	15,650	—
69	▼ 2	qwen-max-0919開源Alibaba	1,318	±6	16,478	—
70	—	athene-v2-chat開源NexusFlow	1,314	±5	24,739	—
71	—	mistral-large-2407開源Mistral · 另有 1 個檔位	1,314	±4	45,459	—
72	—	olmo-3-32b-think開源Ai2	1,306	±8	6,023	—
73	—	athene-70b-0725開源NexusFlow	1,306	±6	19,621	—
74	▲ 1	gemma-3-4b-it開源Google	1,303	±9	4,171	—
75	▼ 1	mistral-small-3.1-24b-instruct-2503開源Mistral	1,303	±4	33,731	—
76	—	qwen2.5-72b-instruct開源Alibaba	1,302	±4	39,406	—
77	—	llama-3.1-nemotron-70b-instruct開源NVIDIA	1,299	±8	7,140	—
78	—	llama-3.1-70b-instruct開源Meta	1,293	±4	55,240	—
79	—	jamba-1.5-large開源AI21 Labs	1,288	±7	8,662	—
80	—	gemma-2-27b-it開源Google	1,288	±3	75,754	—
81	—	ibm-granite-h-small開源IBM	1,287	±8	5,783	—
82	—	llama-3.1-tulu-3-70b開源Ai2	1,286	±10	2,846	—
83	▲ 1	olmo-3.1-32b-think開源Ai2	1,286	±7	8,575	—
84	▼ 1	llama-3.1-nemotron-51b-instruct開源NVIDIA	1,286	±10	3,749	—
85	—	gemma-2-9b-it-simpo開源Princeton	1,279	±7	10,072	—
86	—	nemotron-4-340b-instruct開源NVIDIA	1,277	±5	19,659	—
87	—	command-r-plus-08-2024開源Cohere	1,276	±7	9,866	—
88	—	llama-3-70b-instruct開源Meta	1,275	±4	156,876	—
89	—	mistral-small-24b-instruct-2501開源Mistral	1,274	±6	14,681	—
90	—	qwen2.5-coder-32b-instruct開源Alibaba	1,270	±8	5,432	—
91	—	c4ai-aya-expanse-32b開源Cohere	1,267	±5	27,124	—
92	—	gemma-2-9b-it開源Google	1,265	±4	54,611	—
93	—	deepseek-coder-v2開源DeepSeek	1,264	±6	15,147	—
94	—	command-r-plus開源Cohere	1,261	±4	77,554	—
95	—	qwen2-72b-instruct開源Alibaba	1,261	±5	37,325	—
96	—	phi-4開源Microsoft	1,256	±5	24,126	—
97	—	olmo-2-0325-32b-instruct開源Ai2	1,251	±11	3,334	—
98	—	command-r-08-2024開源Cohere	1,249	±7	10,140	—
99	—	jamba-1.5-mini開源AI21 Labs	1,239	±7	8,858	—
100	—	ministral-8b-2410開源Mistral	1,237	±9	4,781	—
101	—	qwen1.5-110b-chat開源Alibaba	1,233	±6	26,195	—
102	—	qwen1.5-72b-chat開源Alibaba	1,232	±5	39,302	—
103	—	mixtral-8x22b-instruct-v0.1開源Mistral	1,229	±5	51,416	—
104	—	command-r開源Cohere	1,226	±5	54,036	—
105	—	llama-3-8b-instruct開源Meta	1,223	±4	104,642	—
106	—	c4ai-aya-expanse-8b開源Cohere	1,222	±7	9,818	—
107	—	llama-3.1-tulu-3-8b開源Ai2	1,221	±11	2,896	—
108	—	yi-1.5-34b-chat開源01 AI	1,213	±5	24,146	—
109	—	zephyr-orpo-141b-A35b-v0.1開源HuggingFace	1,212	±11	4,652	—
110	—	llama-3.1-8b-instruct開源Meta	1,211	±4	49,605	—
111	—	granite-3.1-8b-instruct開源IBM	1,208	±11	3,090	—
112	—	qwen1.5-32b-chat開源Alibaba	1,203	±6	21,741	—
113	—	gemma-2-2b-it開源Google	1,199	±4	46,616	—
114	—	phi-3-medium-4k-instruct開源Microsoft	1,197	±5	25,055	—
115	—	mixtral-8x7b-instruct-v0.1開源Mistral	1,196	±4	73,503	—
116	—	dbrx-instruct-preview開源Databricks	1,194	±6	32,191	—
117	—	internlm2_5-20b-chat開源InternLM	1,191	±7	9,901	—
118	—	qwen1.5-14b-chat開源Alibaba	1,190	±7	17,839	—
119	—	wizardlm-70b開源Microsoft	1,184	±9	8,214	—
120	—	deepseek-llm-67b-chat開源DeepSeek	1,184	±12	4,932	—
121	—	yi-34b-chat開源01 AI	1,183	±7	15,483	—
122	—	openchat-3.5-0106開源OpenChat · 另有 1 個檔位	1,181	±8	12,637	—
123	—	granite-3.0-8b-instruct開源IBM	1,181	±9	6,638	—
124	—	gemma-1.1-7b-it開源Google	1,180	±6	23,893	—
125	—	snowflake-arctic-instruct開源Snowflake	1,179	±6	32,832	—
126	—	granite-3.1-2b-instruct開源IBM	1,178	±11	3,188	—
127	—	tulu-2-dpo-70b開源AllenAI/UW	1,177	±10	6,535	—
128	—	openhermes-2.5-mistral-7b開源NousResearch	1,174	±10	5,006	—
129	—	vicuna-33b開源LMSYS	1,172	±6	22,479	—
130	▲ 1	starling-lm-7b-beta開源Nexusflow	1,171	±7	16,056	—
131	▼ 1	phi-3-small-8k-instruct開源Microsoft	1,170	±6	17,766	—
132	—	llama-2-70b-chat開源Meta	1,170	±6	38,492	—
133	—	starling-lm-7b-alpha開源UC Berkeley	1,167	±8	10,224	—
134	—	llama-3.2-3b-instruct開源Meta	1,166	±8	7,936	—
135	—	nous-hermes-2-mixtral-8x7b-dpo開源NousResearch	1,164	±12	3,777	—
136	—	granite-3.0-2b-instruct開源IBM	1,155	±8	6,837	—
137	—	llama2-70b-steerlm-chat開源NVIDIA	1,155	±13	3,585	—
138	▲ 1	solar-10.7b-instruct-v1.0開源Upstage AI	1,152	±13	4,155	—
139	▼ 1	dolphin-2.2.1-mistral-7b開源Cognitive Computations	1,151	±15	1,679	—
140	—	mpt-30b-chat開源MosaicML	1,149	±12	2,572	—
141	—	mistral-7b-instruct-v0.2開源Mistral	1,149	±7	19,402	—
142	—	wizardlm-13b開源Microsoft	1,148	±9	7,044	—
143	—	falcon-180b-chat開源TII	1,146	±17	1,295	—
144	—	qwen1.5-7b-chat開源Alibaba	1,143	±10	4,737	—
145	—	phi-3-mini-4k-instruct-june-2024開源Microsoft	1,142	±6	12,297	—
146	—	llama-2-13b-chat開源Meta	1,141	±7	19,174	—
147	—	vicuna-13b開源LMSYS	1,140	±7	19,367	—
148	—	qwen-14b-chat開源Alibaba	1,138	±11	4,964	—
149	—	codellama-34b-instruct開源Meta	1,136	±9	7,366	—
150	—	gemma-7b-it開源Google	1,136	±10	8,925	—
151	—	zephyr-7b-beta開源HuggingFace	1,130	±9	11,118	—
152	—	phi-3-mini-128k-instruct開源Microsoft	1,128	±7	20,685	—
153	—	phi-3-mini-4k-instruct開源Microsoft	1,128	±6	20,118	—
154	—	guanaco-33b開源UW	1,126	±12	2,921	—
155	—	zephyr-7b-alpha開源HuggingFace	1,126	±16	1,785	—
156	—	stripedhyena-nous-7b開源Together AI	1,120	±11	5,182	—
157	—	codellama-70b-instruct開源Meta	1,118	±18	1,143	—
158	—	vicuna-7b開源LMSYS	1,114	±9	6,923	—
159	—	gemma-1.1-2b-it開源Google	1,114	±8	10,854	—
160	—	smollm2-1.7b-instruct開源HuggingFace	1,114	±14	2,199	—
161	—	llama-3.2-1b-instruct開源Meta	1,111	±8	8,045	—
162	—	mistral-7b-instruct開源Mistral	1,109	±9	8,977	—
163	—	llama-2-7b-chat開源Meta	1,107	±7	14,148	—
164	—	gemma-2b-it開源Google	1,091	±12	4,780	—
165	—	qwen1.5-4b-chat開源Alibaba	1,089	±9	7,597	—
166	—	olmo-7b-instruct開源Ai2	1,074	±11	6,328	—
167	—	koala-13b開源UC Berkeley	1,070	±10	6,965	—
168	—	alpaca-13b開源Stanford	1,067	±12	5,745	—
169	—	gpt4all-13b-snoozy開源Nomic AI	1,065	±15	1,743	—
170	—	mpt-7b-chat開源MosaicML	1,061	±12	3,924	—
171	—	chatglm3-6b開源Tsinghua	1,055	±12	4,658	—
172	—	RWKV-4-Raven-14B開源RWKV	1,040	±11	4,845	—
173	—	chatglm2-6b開源Tsinghua	1,023	±14	2,658	—
174	—	oasst-pythia-12b開源OpenAssistant	1,021	±11	6,310	—
175	—	chatglm-6b開源Tsinghua	995	±13	4,914	—
176	—	fastchat-t5-3b開源LMSYS	990	±12	4,203	—
177	—	dolly-v2-12b開源Databricks	979	±14	3,412	—
178	—	llama-13b開源Meta	971	±16	2,391	—
179	—	stablelm-tuned-alpha-7b開源Stability AI	952	±13	3,287	—

其餘 129 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