模型排行

排名	变化	模型	智力	价格 $/M	速度 tok/s	首字延迟
1	—	Kimi K2.6开源Kimi	53.9	$1.7	78	2.5 秒
2	—	MiMo-V2.5-Pro开源Xiaomi	53.8	$1.5	57	2.7 秒
3	—	DeepSeek V4 Pro (Reasoning, Max Effort)开源DeepSeek	51.5	$2.2	30	1.9 秒
4	—	GLM-5.1 (Reasoning)开源Z AI	51.4	$2.2	56	1.4 秒
5	—	DeepSeek V4 Pro (Reasoning, High Effort)开源DeepSeek	49.8	$2.2	30	1.9 秒
6	—	GLM-5 (Reasoning)开源Z AI	49.8	$1.6	66	1.6 秒
7	—	MiniMax-M2.7开源MiniMax	49.6	$0.52	50	2.1 秒
8	—	MiMo-V2.5开源Xiaomi	49.0	$0.72	94	3.3 秒
9	—	Kimi K2.5 (Reasoning)开源Kimi	46.8	$1.2	38	3.0 秒
10	—	DeepSeek V4 Flash (Reasoning, Max Effort)开源DeepSeek	46.5	$0.18	110	1.2 秒
11	—	DeepSeek V4 Flash (Reasoning, High Effort)开源DeepSeek	46.0	$0.18	—	—
12	—	Qwen3.6 27B (Reasoning)开源Alibaba	45.8	$1.4	61	3.8 秒
13	—	Qwen3.5 397B A17B (Reasoning)开源Alibaba	45.0	$1.4	52	2.6 秒
14	—	GLM-5.1 (Non-reasoning)开源Z AI	43.8	$2.2	45	1.8 秒
15	—	Qwen3.6 35B A3B (Reasoning)开源Alibaba	43.5	$0.56	191	2.5 秒
16	—	Kimi K2.6 (Non-reasoning)开源Kimi	42.9	$1.7	55	2.2 秒
17	—	GLM-4.7 (Reasoning)开源Z AI	42.1	$1.0	90	1.3 秒
18	—	Qwen3.5 27B (Reasoning)开源Alibaba	42.1	$0.83	86	5.7 秒
19	—	MiniMax-M2.5开源MiniMax	41.9	$0.52	89	1.8 秒
20	—	Hy3-preview (Reasoning)开源Tencent	41.9	$0.20	104	4.0 秒
21	—	DeepSeek V3.2 (Reasoning)开源DeepSeek	41.7	$0.34	—	—
22	—	Qwen3.5 122B A10B (Reasoning)开源Alibaba	41.6	$1.1	132	2.6 秒
23	—	MiMo-V2-Flash (Feb 2026)开源Xiaomi	41.5	$0.15	136	2.2 秒
24	—	Kimi K2 Thinking开源Kimi	40.9	$1.1	103	2.6 秒
25	—	GLM-5 (Non-reasoning)开源Z AI	40.6	$1.6	45	2.3 秒
26	—	Qwen3.5 397B A17B (Non-reasoning)开源Alibaba	40.1	$1.4	52	2.7 秒
27	—	MiniMax-M2.1开源MiniMax	39.4	$0.52	97	1.7 秒
28	—	DeepSeek V4 Pro (Non-reasoning)开源DeepSeek	39.3	$2.2	30	1.9 秒
29	—	MiMo-V2-Flash (Reasoning)开源Xiaomi	39.2	$0.15	142	2.3 秒
30	—	Mistral Medium 3.5开源Mistral	39.2	$3.0	152	1.7 秒
31	—	Gemma 4 31B (Reasoning)开源Google	39.2	$0.000	35	1.7 秒
32	新	Ring-2.6-1T开源InclusionAI	38.5	—	—	—
33	▼ 1	Step 3.5 Flash开源StepFun	37.8	$0.15	200	1.1 秒
34	▼ 1	Kimi K2.5 (Non-reasoning)开源Kimi	37.3	$1.2	35	3.0 秒
35	▼ 1	Qwen3.5 27B (Non-reasoning)开源Alibaba	37.2	$0.84	89	5.8 秒
36	新	Command A+开源Cohere	37.2	$0.000	267	0.3 秒
37	▼ 2	Qwen3.6 27B (Non-reasoning)开源Alibaba	37.1	$1.4	65	3.8 秒
38	▼ 2	Qwen3.5 35B A3B (Reasoning)开源Alibaba	37.1	$0.69	175	2.3 秒
39	▼ 2	DeepSeek V4 Flash (Non-reasoning)开源DeepSeek	36.5	$0.18	115	1.2 秒
40	▼ 2	MiniMax-M2开源MiniMax	36.1	$0.52	99	1.8 秒
41	▼ 2	NVIDIA Nemotron 3 Super 120B A12B (Reasoning)开源NVIDIA	36.0	$0.41	177	1.2 秒
42	▼ 2	Qwen3.5 122B A10B (Non-reasoning)开源Alibaba	35.9	$1.1	129	2.4 秒
43	▼ 2	MiMo-V2.5-Pro (Non-reasoning)开源Xiaomi	35.6	$1.5	61	2.6 秒
44	▼ 2	GLM-4.7 (Non-reasoning)开源Z AI	34.2	$1.0	103	1.4 秒
45	▼ 2	DeepSeek V3.1 Terminus (Reasoning)开源DeepSeek	33.9	$1.9	—	—
46	▼ 2	Hy3-preview (Non-reasoning)开源Tencent	33.7	$0.20	104	4.1 秒
47	▼ 2	Ling-2.6-1T开源InclusionAI	33.6	$0.85	—	—
48	新	gpt-oss-120b (high)开源OpenAI	33.3	$0.26	253	0.9 秒
49	▼ 2	DeepSeek V3.2 Exp (Reasoning)开源DeepSeek	32.9	$0.31	—	—
50	▼ 2	GLM-4.6 (Reasoning)开源Z AI	32.5	$0.96	40	3.0 秒
51	▼ 2	Qwen3.5 9B (Reasoning)开源Alibaba	32.4	$0.11	63	0.7 秒
52	▼ 2	Gemma 4 31B (Non-reasoning)开源Google	32.3	$0.21	20	2.1 秒
53	▼ 2	K-EXAONE (Reasoning)开源LG AI Research	32.1	—	—	—
54	▼ 2	DeepSeek V3.2 (Non-reasoning)开源DeepSeek	32.1	$0.78	—	—
55	▼ 2	Trinity Large Thinking开源Arcee AI	31.9	$0.40	125	1.1 秒
56	▼ 2	Qwen3.6 35B A3B (Non-reasoning)开源Alibaba	31.5	$0.84	168	2.5 秒
57	▼ 2	Gemma 4 26B A4B (Reasoning)开源Google	31.2	$0.20	—	—
58	▼ 2	Kimi K2 0905开源Kimi	30.9	$1.1	17	2.9 秒
59	▼ 2	Qwen3.5 35B A3B (Non-reasoning)开源Alibaba	30.7	$0.69	174	2.2 秒
60	▼ 2	MiMo-V2-Flash (Non-reasoning)开源Xiaomi	30.3	$0.15	133	2.3 秒
61	▼ 2	GLM-4.6 (Non-reasoning)开源Z AI	30.2	$1.0	35	3.7 秒
62	▼ 2	EXAONE 4.5 33B开源LG AI Research	30.2	—	—	—
63	▼ 2	GLM-4.7-Flash (Reasoning)开源Z AI	30.1	$0.15	83	1.3 秒
64	▼ 2	Qwen3 235B A22B 2507 (Reasoning)开源Alibaba	29.5	$0.84	55	2.9 秒
65	▼ 2	DeepSeek V3.2 Speciale开源DeepSeek	29.4	—	—	—
66	▼ 2	DeepSeek V3.1 Terminus (Non-reasoning)开源DeepSeek	28.5	$0.45	—	—
67	▼ 2	DeepSeek V3.2 Exp (Non-reasoning)开源DeepSeek	28.4	$0.31	—	—
68	▼ 2	Nemotron Cascade 2 30B A3B开源NVIDIA	28.4	—	—	—
69	▼ 2	Qwen3 Coder Next开源Alibaba	28.3	$0.56	93	1.7 秒
70	▼ 2	DeepSeek V3.1 (Non-reasoning)开源DeepSeek	28.1	$0.83	—	—
71	▼ 2	Mistral Small 4 (Reasoning)开源Mistral	27.8	$0.26	135	0.8 秒
72	▼ 2	DeepSeek V3.1 (Reasoning)开源DeepSeek	27.7	$0.87	—	—
73	▼ 2	Qwen3 VL 235B A22B (Reasoning)开源Alibaba	27.6	$2.2	35	4.2 秒
74	▼ 2	Qwen3.5 9B (Non-reasoning)开源Alibaba	27.3	—	—	—
75	▼ 2	Gemma 4 26B A4B (Non-reasoning)开源Google	27.1	$0.20	73	1.3 秒
76	▼ 2	Qwen3.5 4B (Reasoning)开源Alibaba	27.1	$0.060	165	0.5 秒
77	▼ 2	DeepSeek R1 0528 (May '25)开源DeepSeek	27.1	$2.1	—	—
78	▼ 2	Qwen3 Next 80B A3B (Reasoning)开源Alibaba	26.7	$1.9	146	2.2 秒
79	▼ 2	GLM-4.5 (Reasoning)开源Z AI	26.4	$1.0	50	3.6 秒
80	▼ 2	Kimi K2开源Kimi	26.3	$1.0	27	2.9 秒
81	▼ 2	Ling 2.6 Flash开源InclusionAI	26.2	$0.15	—	—
82	▼ 2	Seed-OSS-36B-Instruct开源ByteDance Seed	25.2	$0.30	36	3.5 秒
83	▼ 2	Qwen3 235B A22B 2507 Instruct开源Alibaba	25.0	$0.36	61	2.4 秒
84	▼ 2	Qwen3 Coder 480B A35B Instruct开源Alibaba	24.8	$0.68	58	3.0 秒
85	▼ 2	Qwen3 VL 32B (Reasoning)开源Alibaba	24.7	$2.6	90	2.7 秒
86	▼ 2	gpt-oss-20B (high)开源OpenAI	24.5	$0.088	239	0.8 秒
87	新	gpt-oss-120b (low)开源OpenAI	24.5	$0.26	281	0.8 秒
88	▼ 2	MiniMax M1 80k开源MiniMax	24.4	$0.96	—	—
89	▼ 2	NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)开源NVIDIA	24.3	$0.096	150	1.4 秒
90	▼ 2	LongCat Flash Lite开源LongCat	23.9	$0.000	114	8.6 秒
91	▼ 2	GLM-4.6V (Reasoning)开源Z AI	23.4	$0.45	46	4.7 秒
92	▼ 2	K-EXAONE (Non-reasoning)开源LG AI Research	23.4	—	—	—
93	▼ 2	GLM-4.5-Air开源Z AI	23.2	$0.37	61	3.1 秒
94	▼ 2	Mistral Large 3开源Mistral	22.8	$0.75	48	1.1 秒
95	▼ 2	Ring-1T开源InclusionAI	22.8	—	—	—
96	▼ 2	Qwen3.5 4B (Non-reasoning)开源Alibaba	22.6	$0.060	161	0.5 秒
97	▼ 2	Qwen3 30B A3B 2507 (Reasoning)开源Alibaba	22.4	$0.67	139	2.4 秒
98	▼ 2	DeepSeek V3 0324开源DeepSeek	22.3	$1.2	—	—
99	▼ 2	INTELLECT-3开源Prime Intellect	22.2	—	—	—
100	▼ 2	GLM-4.7-Flash (Non-reasoning)开源Z AI	22.1	$0.15	134	1.8 秒
101	▼ 2	Devstral 2开源Mistral	22.0	$0.000	47	1.3 秒
102	▼ 2	Solar Open 100B (Reasoning)开源Upstage	21.7	—	—	—
103	▼ 2	Nemotron 3 Nano Omni 30B A3B Reasoning开源NVIDIA	21.4	$0.13	301	1.0 秒
104	▼ 2	MiniMax M1 40k开源MiniMax	20.9	—	—	—
105	▼ 2	gpt-oss-20B (low)开源OpenAI	20.8	$0.095	232	0.8 秒
106	▼ 2	Qwen3 VL 235B A22B Instruct开源Alibaba	20.8	$0.70	50	2.8 秒
107	▼ 2	Qwen3 Next 80B A3B Instruct开源Alibaba	20.1	$0.88	146	2.4 秒
108	▼ 2	Qwen3 Coder 30B A3B Instruct开源Alibaba	20.0	$0.35	94	2.7 秒
109	▼ 2	Qwen3 235B A22B (Reasoning)开源Alibaba	19.8	$2.6	60	2.9 秒
110	▼ 2	QwQ 32B开源Alibaba	19.7	$0.75	29	2.1 秒
111	▼ 2	Qwen3 VL 30B A3B (Reasoning)开源Alibaba	19.7	$0.34	111	2.4 秒
112	▼ 2	Devstral Small 2开源Mistral	19.5	$0.000	50	1.2 秒
113	▼ 2	Ling-1T开源InclusionAI	19.0	—	—	—
114	▼ 2	DeepSeek R1 (Jan '25)开源DeepSeek	18.8	$2.4	—	—
115	▼ 2	Gemma 4 E4B (Reasoning)开源Google	18.8	—	—	—
116	▼ 2	Llama Nemotron Super 49B v1.5 (Reasoning)开源NVIDIA	18.7	$0.18	50	1.3 秒
117	▼ 2	Mistral Small 4 (Non-reasoning)开源Mistral	18.6	$0.26	137	0.9 秒
118	▼ 2	Llama 3.3 Nemotron Super 49B v1 (Reasoning)开源NVIDIA	18.5	—	—	—
119	▼ 2	Llama 4 Maverick开源Meta	18.4	$0.48	108	1.1 秒
120	▼ 2	Qwen3 4B 2507 (Reasoning)开源Alibaba	18.2	—	—	—
121	▼ 2	Magistral Small 1.2开源Mistral	18.2	$0.75	105	0.8 秒
122	▼ 2	Sarvam 105B (high)开源Sarvam	18.2	$0.000	112	2.2 秒
123	▼ 2	Devstral Small (May '25)开源Mistral	18.0	—	—	—
124	▼ 2	Llama 3.1 Instruct 405B开源Meta	17.4	$3.7	53	2.4 秒
125	▼ 2	Qwen3 VL 32B Instruct开源Alibaba	17.2	$1.2	78	2.6 秒
126	▼ 2	DeepSeek R1 Distill Qwen 32B开源DeepSeek	17.2	—	—	—
127	▼ 2	GLM-4.6V (Non-reasoning)开源Z AI	17.1	$0.45	38	5.8 秒
128	▼ 2	Qwen3 235B A22B (Non-reasoning)开源Alibaba	17.0	$0.79	61	2.8 秒
129	▼ 2	Magistral Small 1开源Mistral	16.8	—	—	—
130	▼ 2	EXAONE 4.0 32B (Reasoning)开源LG AI Research	16.7	—	—	—
131	▼ 2	Qwen3 VL 8B (Reasoning)开源Alibaba	16.7	$0.66	108	2.5 秒
132	▼ 2	Qwen3 32B (Reasoning)开源Alibaba	16.5	$0.28	86	2.7 秒
133	▼ 2	DeepSeek V3 (Dec '24)开源DeepSeek	16.5	$0.52	—	—
134	▼ 2	DeepSeek R1 0528 Qwen3 8B开源DeepSeek	16.4	—	—	—
135	▼ 2	Qwen3.5 2B (Reasoning)开源Alibaba	16.3	$0.040	—	—
136	▼ 2	Qwen3 14B (Reasoning)开源Alibaba	16.2	$0.73	58	2.8 秒
137	▼ 2	Nanbeige4.1-3B开源Nanbeige	16.1	—	—	—
138	▼ 2	Qwen3 VL 30B A3B Instruct开源Alibaba	16.0	$0.30	111	2.2 秒
139	▼ 2	Ministral 3 14B开源Mistral	16.0	$0.20	67	0.7 秒
140	▼ 2	DeepSeek R1 Distill Llama 70B开源DeepSeek	16.0	$0.79	42	1.7 秒
141	▼ 2	DeepSeek R1 Distill Qwen 14B开源DeepSeek	15.8	—	—	—
142	▼ 2	Falcon-H1R-7B开源TII UAE	15.8	—	—	—
143	▼ 2	Ling-flash-2.0开源InclusionAI	15.7	$0.25	82	2.4 秒
144	▼ 2	Qwen3 Omni 30B A3B (Reasoning)开源Alibaba	15.6	$0.43	86	2.1 秒
145	▼ 2	Qwen2.5 Instruct 72B开源Alibaba	15.6	$0.37	28	3.6 秒
146	▼ 2	Step3 VL 10B开源StepFun	15.5	—	—	—
147	▼ 2	Qwen3 30B A3B (Reasoning)开源Alibaba	15.3	$0.18	60	2.5 秒
148	▼ 2	Devstral Small (Jul '25)开源Mistral	15.2	$0.15	223	0.7 秒
149	▼ 2	Gemma 4 E2B (Reasoning)开源Google	15.2	—	—	—
150	▼ 2	QwQ 32B-Preview开源Alibaba	15.2	—	—	—
151	▼ 2	GLM-4.5V (Reasoning)开源Z AI	15.1	$0.90	38	2.2 秒
152	▼ 2	Mistral Large 2 (Nov '24)开源Mistral	15.1	$3.0	32	2.0 秒
153	▼ 2	Mistral Small 3.2开源Mistral	15.1	$0.13	98	0.7 秒
154	▼ 2	Llama 3.1 Nemotron Ultra 253B v1 (Reasoning)开源NVIDIA	15.0	$0.90	41	2.6 秒
155	▼ 2	Qwen3 30B A3B 2507 Instruct开源Alibaba	15.0	$0.21	105	2.1 秒
156	▼ 2	ERNIE 4.5 300B A47B开源Baidu	15.0	$0.49	23	3.5 秒
157	▼ 2	NVIDIA Nemotron Nano 12B v2 VL (Reasoning)开源NVIDIA	14.9	$0.30	—	—
158	▼ 2	Ministral 3 8B开源Mistral	14.8	$0.15	85	0.7 秒
159	▼ 2	Gemma 4 E4B (Non-reasoning)开源Google	14.8	—	—	—
160	▼ 2	NVIDIA Nemotron Nano 9B V2 (Reasoning)开源NVIDIA	14.8	$0.070	115	0.6 秒
161	▼ 2	Granite 4.1 30B开源IBM	14.7	—	—	—
162	▼ 2	NVIDIA Nemotron 3 Nano 4B开源NVIDIA	14.7	—	—	—
163	▼ 2	Qwen3.5 2B (Non-reasoning)开源Alibaba	14.7	$0.040	327	0.4 秒
164	▼ 2	Llama Nemotron Super 49B v1.5 (Non-reasoning)开源NVIDIA	14.6	$0.18	49	1.3 秒
165	▼ 2	Qwen3 32B (Non-reasoning)开源Alibaba	14.5	$0.26	89	2.7 秒
166	▼ 2	Llama 3.3 Instruct 70B开源Meta	14.5	$0.62	80	1.7 秒
167	▼ 2	Mistral Small 3.1开源Mistral	14.5	$0.14	140	0.7 秒
168	▼ 2	Llama 3.1 Nemotron Nano 4B v1.1 (Reasoning)开源NVIDIA	14.4	—	—	—
169	▼ 2	Kimi Linear 48B A3B Instruct开源Kimi	14.4	—	—	—
170	▼ 2	Llama 3.3 Nemotron Super 49B v1 (Non-reasoning)开源NVIDIA	14.3	—	—	—
171	▼ 2	Qwen3 VL 8B Instruct开源Alibaba	14.3	$0.31	120	2.3 秒
172	▼ 2	Qwen3 4B (Reasoning)开源Alibaba	14.2	$0.40	95	2.4 秒
173	▼ 2	Ring-flash-2.0开源InclusionAI	14.0	$0.25	—	—
174	▼ 2	Pixtral Large开源Mistral	14.0	$3.0	52	1.7 秒
175	▼ 2	Grok 2 (Dec '24)开源xAI	13.9	—	—	—
176	▼ 2	Qwen3 VL 4B (Reasoning)开源Alibaba	13.7	—	—	—
177	▼ 2	Llama 4 Scout开源Meta	13.5	$0.29	119	0.9 秒
178	▼ 2	Command A开源Cohere	13.5	$4.4	37	2.3 秒
179	▼ 2	Llama 3.1 Nemotron Instruct 70B开源NVIDIA	13.4	$1.2	287	0.5 秒
180	▼ 2	Qwen2.5 Instruct 32B开源Alibaba	13.2	—	—	—
181	▼ 2	Qwen3 8B (Reasoning)开源Alibaba	13.2	$0.37	39	3.7 秒
182	▼ 2	NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)开源NVIDIA	13.2	$0.088	74	0.5 秒
183	▼ 2	NVIDIA Nemotron Nano 9B V2 (Non-reasoning)开源NVIDIA	13.2	$0.086	128	1.1 秒
184	▼ 2	Mistral Large 2 (Jul '24)开源Mistral	13.0	$3.0	—	—
185	▼ 2	Qwen3 4B 2507 Instruct开源Alibaba	12.9	—	—	—
186	▼ 2	Qwen2.5 Coder Instruct 32B开源Alibaba	12.9	—	—	—
187	▼ 2	Qwen3 14B (Non-reasoning)开源Alibaba	12.8	$0.38	58	2.8 秒
188	▼ 2	GLM-4.5V (Non-reasoning)开源Z AI	12.7	$0.90	33	25 秒
189	▼ 2	Mistral Small 3开源Mistral	12.7	$0.10	126	0.8 秒
190	▼ 2	MiniCPM-V 4.6 1.3B开源OpenBMB	12.7	—	—	—
191	▼ 2	Qwen3 30B A3B (Non-reasoning)开源Alibaba	12.5	$0.13	68	2.4 秒
192	▼ 2	DeepSeek-V2.5 (Dec '24)开源DeepSeek	12.5	—	—	—
193	▼ 2	Qwen3 4B (Non-reasoning)开源Alibaba	12.5	$0.19	96	2.5 秒
194	▼ 2	Llama 3.1 Instruct 70B开源Meta	12.5	$0.56	32	1.9 秒
195	▼ 2	Granite 4.1 8B开源IBM	12.4	$0.063	114	0.8 秒
196	▼ 2	Sarvam 30B (high)开源Sarvam	12.3	$0.000	262	1.9 秒
197	▼ 2	DeepSeek-V2.5开源DeepSeek	12.3	—	—	—
198	▼ 2	DeepSeek R1 Distill Llama 8B开源DeepSeek	12.1	—	—	—
199	▼ 2	Gemma 4 E2B (Non-reasoning)开源Google	12.1	—	—	—
200	▼ 2	R1 1776开源Perplexity	12.0	—	—	—
201	▼ 2	Llama 3.2 Instruct 90B (Vision)开源Meta	11.9	$1.4	58	1.3 秒
201	▼ 2	Solar Mini开源Upstage	11.9	$0.15	—	—
203	▼ 2	Llama 3.1 Instruct 8B开源Meta	11.8	$0.10	155	0.8 秒
204	▼ 2	Grok-1开源xAI	11.7	—	—	—
205	▼ 2	Qwen2 Instruct 72B开源Alibaba	11.7	—	—	—
206	▼ 2	EXAONE 4.0 32B (Non-reasoning)开源LG AI Research	11.7	—	—	—
207	▼ 2	Ministral 3 3B开源Mistral	11.2	$0.10	150	0.5 秒
208	▼ 2	Jamba 1.7 Large开源AI21 Labs	10.9	$3.5	57	1.5 秒
209	▼ 2	Granite 4.0 H Small开源IBM	10.8	$0.11	388	10 秒
210	▼ 2	Jamba 1.5 Large开源AI21 Labs	10.7	$3.5	—	—
211	▼ 2	Qwen3 Omni 30B A3B Instruct开源Alibaba	10.7	$0.43	96	2.0 秒
212	▼ 2	Qwen3 8B (Non-reasoning)开源Alibaba	10.6	$0.19	40	3.8 秒
213	▼ 2	DeepSeek-Coder-V2开源DeepSeek	10.6	—	—	—
214	▼ 2	Jamba 1.6 Large开源AI21 Labs	10.6	$3.5	58	1.5 秒
215	▼ 2	Qwen3.5 0.8B (Reasoning)开源Alibaba	10.5	$0.020	—	—
216	▼ 2	LFM2 24B A2B开源Liquid AI	10.5	$0.053	209	0.6 秒
217	▼ 2	Phi-4开源Microsoft	10.4	$0.22	18	2.5 秒
218	▼ 2	Gemma 3 27B Instruct开源Google	10.3	$0.14	—	—
219	▼ 2	Mistral Small (Sep '24)开源Mistral	10.2	$0.30	133	0.8 秒
220	▼ 2	Phi-3 Mini Instruct 3.8B开源Microsoft	10.1	—	—	—
221	▼ 2	NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)开源NVIDIA	10.1	$0.30	236	1.1 秒
222	▼ 2	Gemma 3n E4B Instruct Preview (May '25)开源Google	10.1	—	—	—
223	▼ 2	Phi-4 Multimodal Instruct开源Microsoft	10.0	$0.000	16	0.9 秒
224	▼ 2	Qwen2.5 Coder Instruct 7B开源Alibaba	10.0	—	—	—
225	▼ 2	Qwen3.5 0.8B (Non-reasoning)开源Alibaba	9.9	$0.020	117	0.4 秒
226	▼ 2	Mixtral 8x22B Instruct开源Mistral	9.8	—	—	—
227	▼ 2	Llama 2 Chat 7B开源Meta	9.7	$0.10	—	—
228	▼ 2	Llama 3.2 Instruct 3B开源Meta	9.7	$0.15	51	1.2 秒
229	▼ 2	Jamba Reasoning 3B开源AI21 Labs	9.6	—	—	—
230	▼ 2	Qwen3 VL 4B Instruct开源Alibaba	9.6	—	—	—
231	▼ 2	Qwen1.5 Chat 110B开源Alibaba	9.5	—	—	—
232	▼ 2	Reka Flash 3开源Reka AI	9.5	$0.35	—	—
233	▼ 2	Ling-mini-2.0开源InclusionAI	9.2	—	—	—
234	▼ 2	DeepSeek R1 Distill Qwen 1.5B开源DeepSeek	9.1	—	—	—
235	▼ 2	DeepSeek-V2-Chat开源DeepSeek	9.1	—	—	—
236	▼ 2	Llama 3 Instruct 70B开源Meta	8.9	$1.2	—	—
237	▼ 2	Arctic Instruct开源Snowflake	8.8	—	—	—
237	▼ 2	Qwen Chat 72B开源Alibaba	8.8	—	—	—
239	▼ 2	Gemma 3 12B Instruct开源Google	8.8	$0.14	—	—
240	▼ 2	Llama 3.2 Instruct 11B (Vision)开源Meta	8.7	$0.25	48	0.8 秒
241	▼ 2	Granite 4.1 3B开源IBM	8.5	—	—	—
242	▼ 2	DeepSeek Coder V2 Lite Instruct开源DeepSeek	8.5	—	—	—
243	▼ 2	Sarvam M (Reasoning)开源Sarvam	8.4	$0.000	—	—
244	▼ 2	Phi-4 Mini Instruct开源Microsoft	8.4	$0.000	—	—
245	▼ 2	DeepSeek LLM 67B Chat (V1)开源DeepSeek	8.4	—	—	—
245	▼ 2	Llama 2 Chat 70B开源Meta	8.4	—	—	—
247	▼ 2	Llama 2 Chat 13B开源Meta	8.4	—	—	—
248	▼ 2	Command-R+ (Apr '24)开源Cohere	8.3	$6.0	—	—
249	▼ 2	OpenChat 3.5 (1210)开源OpenChat	8.3	—	—	—
250	▼ 2	DBRX Instruct开源Databricks	8.3	—	—	—
251	▼ 2	Exaone 4.0 1.2B (Reasoning)开源LG AI Research	8.3	—	—	—
252	▼ 2	Exaone 4.0 1.2B (Non-reasoning)开源LG AI Research	8.1	—	—	—
253	▼ 2	LFM2.5-1.2B-Thinking开源Liquid AI	8.1	—	—	—
254	▼ 2	Jamba 1.7 Mini开源AI21 Labs	8.1	—	—	—
255	▼ 2	LFM2 2.6B开源Liquid AI	8.0	$0.000	—	—
256	▼ 2	LFM2.5-1.2B-Instruct开源Liquid AI	8.0	$0.000	—	—
257	▼ 2	Jamba 1.5 Mini开源AI21 Labs	8.0	$0.25	—	—
258	▼ 2	Granite 4.0 H 1B开源IBM	8.0	—	—	—
259	▼ 2	Qwen3 1.7B (Reasoning)开源Alibaba	8.0	$0.40	126	1.9 秒
260	▼ 2	Jamba 1.6 Mini开源AI21 Labs	7.9	$0.25	182	0.8 秒
261	▼ 2	Mixtral 8x7B Instruct开源Mistral	7.7	$0.51	—	—
262	▼ 2	Gemma 3 270M开源Google	7.7	—	—	—
263	▼ 2	Apertus 70B Instruct开源Swiss AI Initiative	7.7	$1.3	—	—
264	▼ 2	Granite 4.0 Micro开源IBM	7.7	—	—	—
265	▼ 2	Command-R (Mar '24)开源Cohere	7.4	$0.75	—	—
265	▼ 2	Llama 65B开源Meta	7.4	—	—	—
265	▼ 2	Mistral 7B Instruct开源Mistral	7.4	$0.21	84	0.7 秒
265	▼ 2	Qwen Chat 14B开源Alibaba	7.4	—	—	—
269	▼ 2	Granite 4.0 1B开源IBM	7.3	—	—	—
270	▼ 2	LFM2 8B A1B开源Liquid AI	7.0	$0.000	—	—
271	▼ 2	Granite 3.3 8B (Non-reasoning)开源IBM	7.0	$0.085	343	27 秒
272	▼ 2	Qwen3 1.7B (Non-reasoning)开源Alibaba	6.8	$0.19	127	1.8 秒
273	▼ 2	Qwen3 0.6B (Reasoning)开源Alibaba	6.5	$0.40	195	1.7 秒
274	▼ 2	Llama 3 Instruct 8B开源Meta	6.4	$0.070	—	—
275	▼ 2	Gemma 3n E4B Instruct开源Google	6.4	$0.025	57	1.3 秒
276	▼ 2	LFM2 1.2B开源Liquid AI	6.3	$0.000	—	—
277	▼ 2	Gemma 3 4B Instruct开源Google	6.3	$0.050	—	—
278	▼ 2	Llama 3.2 Instruct 1B开源Meta	6.3	$0.050	87	0.9 秒
279	▼ 2	LFM2.5-VL-1.6B开源Liquid AI	6.2	$0.000	—	—
280	▼ 2	Granite 4.0 350M开源IBM	6.1	—	—	—
281	▼ 2	Apertus 8B Instruct开源Swiss AI Initiative	5.9	$0.13	—	—
282	▼ 2	Qwen3 0.6B (Non-reasoning)开源Alibaba	5.7	$0.19	194	1.7 秒
283	▼ 2	Gemma 3 1B Instruct开源Google	5.6	$0.000	—	—
284	▼ 2	Granite 4.0 H 350M开源IBM	5.4	—	—	—
285	▼ 2	Gemma 3n E2B Instruct开源Google	4.8	$0.000	—	—
286	▼ 2	Tiny Aya Global开源Cohere	4.7	$0.000	—	—

其余 236 个模型

智力进步

每家厂商当时最强的模型在 Artificial Analysis 智力指数上的分数，按模型发布的月份画。

涨得最多 · Anthropic +57.2目前最高 · Anthropic 64.6最高与最低相差 18.1 分时间范围 2025年11月 – 2026年11月

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