模型排行

排名	變化	模型	智力	價格 $/M	速度 tok/s	首字延遲
1	—	DeepSeek R1 0528 (May '25)開源DeepSeek	68.3	$0.96	—	—
2	—	MiniMax M1 80k開源MiniMax	63.0	$0.83	—	—
3	—	Qwen3 235B A22B (Reasoning)開源Alibaba	62.3	$2.6	—	—
4	—	MiniMax M1 40k開源MiniMax	61.4	$0.83	—	—
5	—	Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)開源NVIDIA	60.8	$0.90	—	—
6	—	DeepSeek R1 (Jan '25)開源DeepSeek	60.2	$2.0	—	—
7	—	Qwen3 32B (Reasoning)開源Alibaba	59.2	$2.6	—	—
8	—	QwQ 32B開源Alibaba	58.1	$0.33	—	—
9	—	Qwen3 14B (Reasoning)開源Alibaba	55.7	$1.3	—	—
10	—	Qwen3 30B A3B (Reasoning)開源Alibaba	55.6	$0.75	—	—
11	—	Magistral Small開源Mistral	54.7	$0.75	—	—
12	—	DeepSeek V3 0324 (Mar '25)開源DeepSeek	53.2	$0.48	—	—
13	—	DeepSeek R1 0528 Qwen3 8B開源DeepSeek	52.2	$0.065	—	—
14	—	DeepSeek R1 Distill Qwen 32B開源DeepSeek	51.9	$0.30	—	—
15	—	Qwen3 8B (Reasoning)開源Alibaba	51.4	$0.66	—	—
16	—	Llama 3.3 Nemotron Super 49B v1 (Reasoning)開源NVIDIA	51.2	$0.000	—	—
17	—	Llama 4 Maverick開源Meta	50.5	$0.39	—	—
18	—	DeepSeek R1 Distill Qwen 14B開源DeepSeek	49.0	$0.20	—	—
19	—	DeepSeek R1 Distill Llama 70B開源DeepSeek	48.2	$0.72	—	—
20	—	Qwen3 4B (Reasoning)開源Alibaba	47.3	$0.40	—	—
21	—	Reka Flash 3開源Reka AI	47.1	$0.35	—	—
22	—	Qwen3 235B A22B開源Alibaba	47.1	$1.2	—	—
23	—	DeepSeek V3 (Dec '24)開源DeepSeek	45.6	$0.48	—	—
24	—	Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)開源NVIDIA	45.0	$0.000	—	—
25	—	Qwen3 32B開源Alibaba	43.5	$1.2	—	—
26	—	Llama 4 Scout開源Meta	43.0	$0.24	—	—
27	—	QwQ 32B-Preview開源Alibaba	42.6	$0.67	—	—
28	—	Qwen3 30B A3B開源Alibaba	42.6	$0.35	—	—
29	—	Mistral Small 3.2開源Mistral	42.3	$0.15	—	—
30	—	Llama 3.3 Instruct 70B開源Meta	41.1	$0.62	—	—
31	—	Qwen3 14B開源Alibaba	41.0	$0.61	—	—
32	—	Llama 3.1 Instruct 405B開源Meta	40.5	$3.3	—	—
33	—	Qwen2.5 Instruct 72B開源Alibaba	40.4	$0.000	—	—
34	—	MiniMax-Text-01開源MiniMax	40.2	$0.43	—	—
35	—	Phi-4開源Microsoft Azure	40.2	$0.22	—	—
36	—	Command A開源Cohere	40.0	$4.4	—	—
37	—	Llama 3.1 Tulu3 405B開源Allen Institute for AI	39.9	$0.000	—	—
38	—	Llama 3.3 Nemotron Super 49B v1開源NVIDIA	39.3	$0.000	—	—
39	—	Mistral Large 2 (Nov '24)開源Mistral	38.3	$3.0	—	—
40	—	Qwen3 1.7B (Reasoning)開源Alibaba	38.2	$0.40	—	—
41	—	Gemma 3 27B Instruct開源Google	37.6	$0.000	—	—
42	—	Pixtral Large開源Mistral	37.4	$3.0	—	—
43	—	Qwen2.5 Instruct 32B開源Alibaba	37.4	$0.15	—	—
44	—	Llama 3.1 Nemotron Instruct 70B開源NVIDIA	37.3	$0.16	—	—
45	—	Qwen3 8B開源Alibaba	36.7	$0.31	—	—
46	—	Mistral Large 2 (Jul '24)開源Mistral	36.6	$3.0	—	—
47	—	Qwen2.5 Coder Instruct 32B開源Alibaba	36.3	$0.15	—	—
48	—	Llama 3.1 Instruct 70B開源Meta	35.4	$0.76	—	—
49	—	Mistral Small 3.1開源Mistral	35.3	$0.15	—	—
50	—	Mistral Small 3開源Mistral	35.3	$0.15	—	—
51	—	DeepSeek-V2.5 (Dec '24)開源DeepSeek	35.3	$0.18	—	—
52	—	Qwen3 4B開源Alibaba	35.2	$0.19	—	—
53	—	DeepSeek-V2.5開源DeepSeek	34.7	$0.18	—	—
54	—	Devstral開源Mistral	34.1	$0.15	—	—
55	—	DeepSeek R1 Distill Llama 8B開源DeepSeek	34.0	$0.040	—	—
56	—	Gemma 3 12B Instruct開源Google	33.8	$0.063	—	—
57	—	R1 1776開源Perplexity	33.7	$3.5	—	—
58	—	Llama 3.2 Instruct 90B (Vision)開源Meta	33.4	$0.54	—	—
58	—	Solar Mini開源Upstage	33.4	$0.15	—	—
60	—	Qwen2 Instruct 72B開源Alibaba	32.6	$0.000	—	—
61	—	Gemma 2 27B開源Google	31.7	$0.80	—	—
62	—	DeepHermes 3 - Mistral 24B Preview開源Nous Research	30.0	$0.000	—	—
63	—	Jamba 1.5 Large開源AI21 Labs	29.3	$3.5	—	—
64	—	Hermes 3 - Llama-3.1 70B開源Nous Research	29.1	$0.000	—	—
65	—	DeepSeek-Coder-V2開源DeepSeek	29.0	$0.18	—	—
66	—	Jamba 1.6 Large開源AI21 Labs	28.8	$3.5	—	—
67	—	Gemma 3n E4B Instruct開源Google	28.0	$0.025	—	—
68	—	Llama 3 Instruct 70B開源Meta	27.5	$0.75	—	—
69	—	Mistral Small (Sep '24)開源Mistral	27.4	$0.30	—	—
70	—	Gemma 3n E4B Instruct Preview (May '25)開源Google	27.0	$0.000	—	—
71	—	Phi-4 Multimodal Instruct開源Microsoft Azure	26.9	$0.000	—	—
72	—	Qwen2.5 Coder Instruct 7B開源Alibaba	26.7	$0.000	—	—
73	—	Mixtral 8x22B Instruct開源Mistral	26.2	$3.0	—	—
74	—	Phi-4 Mini Instruct開源Microsoft Azure	26.0	$0.000	—	—
75	—	Gemma 3 4B Instruct開源Google	25.4	$0.025	—	—
76	—	Llama 3.2 Instruct 11B (Vision)開源Meta	25.0	$0.10	—	—
76	—	Qwen3 1.7B開源Alibaba	25.0	$0.19	—	—
78	—	Qwen1.5 Chat 110B開源Alibaba	25.0	$0.000	—	—
79	—	Phi-3 Medium Instruct 14B開源Microsoft Azure	24.5	$0.30	—	—
80	—	Llama 3.1 Instruct 8B開源Meta	23.7	$0.10	—	—
81	—	Pixtral 12B (2409)開源Mistral	23.4	$0.15	—	—
82	—	Qwen3 0.6B (Reasoning)開源Alibaba	23.2	$0.40	—	—
83	—	DeepSeek-V2-Chat開源DeepSeek	23.0	$0.18	—	—
84	—	Ministral 8B開源Mistral	22.3	$0.10	—	—
85	—	Gemma 2 9B開源Google	22.2	$0.20	—	—
86	—	Phi-3 Mini Instruct 3.8B開源Microsoft Azure	22.1	$0.000	—	—
87	—	Arctic Instruct開源Snowflake	22.0	$0.000	—	—
87	—	Qwen Chat 72B開源Alibaba	22.0	$1.0	—	—
89	—	Command-R+ (Aug '24)開源Cohere	21.5	$4.4	—	—
90	—	Llama 3 Instruct 8B開源Meta	21.5	$0.058	—	—
91	—	DeepSeek Coder V2 Lite Instruct開源DeepSeek	20.5	$0.000	—	—
92	—	Codestral (May '24)開源Mistral	20.4	$0.30	—	—
93	—	Aya Expanse 32B開源Cohere	20.1	$0.75	—	—
94	—	DeepSeek LLM 67B Chat (V1)開源DeepSeek	20.0	$0.000	—	—
94	—	Llama 2 Chat 70B開源Meta	20.0	$0.000	—	—
96	—	Llama 2 Chat 13B開源Meta	19.9	$0.000	—	—
97	—	Command-R+ (Apr '24)開源Cohere	19.9	$6.0	—	—
98	—	OpenChat 3.5 (1210)開源OpenChat	19.8	$0.000	—	—
99	—	DBRX Instruct開源Databricks	19.8	$0.000	—	—
100	—	Mistral NeMo開源Mistral	19.6	$0.15	—	—
101	—	Llama 3.2 Instruct 3B開源Meta	19.5	$0.035	—	—
102	—	DeepSeek R1 Distill Qwen 1.5B開源DeepSeek	19.2	$0.000	—	—
103	—	Jamba 1.5 Mini開源AI21 Labs	18.4	$0.25	—	—
104	—	Jamba 1.6 Mini開源AI21 Labs	17.7	$0.25	—	—
105	—	Mixtral 8x7B Instruct開源Mistral	17.0	$0.70	—	—
106	—	Qwen3 0.6B開源Alibaba	16.7	$0.19	—	—
107	—	DeepHermes 3 - Llama-3.1 8B Preview開源Nous Research	16.2	$0.000	—	—
108	—	Aya Expanse 8B開源Cohere	16.0	$0.75	—	—
109	—	Command-R (Aug '24)開源Cohere	14.8	$0.26	—	—
110	—	Command-R (Mar '24)開源Cohere	14.7	$0.75	—	—
111	—	Qwen Chat 14B開源Alibaba	14.0	$0.000	—	—
112	—	Codestral-Mamba開源Mistral	13.9	$0.000	—	—
113	—	Gemma 3 1B Instruct開源Google	13.4	$0.000	—	—
114	—	Llama 65B開源Meta	11.0	$0.000	—	—
115	—	Mistral 7B Instruct開源Mistral	10.1	$0.25	—	—
116	—	Llama 3.2 Instruct 1B開源Meta	9.7	$0.053	—	—
117	—	Llama 2 Chat 7B開源Meta	8.3	$0.10	—	—

其餘 67 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