模型排行

排名	变化	模型	得分	置信区间	投票	价格 $/M
1	—	o3-2025-04-16闭源OpenAI	1,443	±5	13,808	—
2	—	chatgpt-4o-latest-20250326闭源OpenAI	1,431	±4	18,302	—
3	—	gpt-4.5-preview-2025-02-27闭源OpenAI	1,425	±4	15,271	—
4	—	claude-opus-4-20250514闭源Anthropic	1,414	±6	10,402	—
5	—	gpt-4.1-2025-04-14闭源OpenAI	1,402	±5	12,421	—
6	—	grok-3-preview-02-24闭源xAI	1,399	±4	20,585	—
7	▲ 1	o4-mini-2025-04-16闭源OpenAI	1,390	±5	12,123	—
8	▼ 1	claude-sonnet-4-20250514闭源Anthropic	1,390	±6	8,087	—
9	—	o1-2024-12-17闭源OpenAI	1,388	±3	29,038	—
10	—	o1-preview闭源OpenAI	1,372	±3	33,177	—
11	▲ 1	claude-3-7-sonnet-20250219闭源Anthropic	1,365	±4	24,525	—
12	▲ 1	gpt-4.1-mini-2025-04-14闭源OpenAI	1,365	±5	11,210	—
13	▼ 2	claude-3-7-sonnet-20250219-thinking-32k闭源Anthropic	1,363	±4	18,682	—
14	▲ 1	gemini-2.0-flash-001闭源Google	1,355	±3	33,324	—
15	▼ 1	claude-3-5-sonnet-20241022闭源Anthropic · 另有 1 个档位	1,355	±2	72,401	—
16	—	o3-mini-high闭源OpenAI	1,355	±4	19,404	—
17	—	gemma-3-27b-it开源Google	1,353	±4	18,934	—
18	—	gemini-2.0-flash-lite-preview-02-05闭源Google	1,339	±3	26,104	—
19	—	o3-mini闭源OpenAI	1,339	±3	31,005	—
20	—	gemini-1.5-pro-002闭源Google	1,337	±3	58,645	—
21	—	gpt-4o-2024-05-13闭源OpenAI	1,332	±2	117,747	—
22	—	gemma-3-12b-it开源Google	1,330	±9	3,976	—
23	—	o1-mini闭源OpenAI	1,325	±2	54,951	—
24	—	gemini-advanced-0514闭源Google	1,323	±3	52,144	—
25	—	llama-3.1-405b-instruct-bf16开源Meta	1,323	±2	43,788	—
26	—	llama-3.1-405b-instruct-fp8开源Meta	1,322	±2	63,038	—
27	—	gpt-4o-2024-08-06闭源OpenAI	1,319	±3	47,973	—
28	—	llama-4-maverick-17b-128e-instruct开源Meta	1,319	±4	13,749	—
29	—	grok-2-2024-08-13闭源xAI	1,318	±2	67,084	—
30	—	llama-3.3-nemotron-49b-super-v1开源NVIDIA	1,315	±11	2,371	—
31	▲ 1	gpt-4-turbo-2024-04-09闭源OpenAI	1,311	±2	102,133	—
32	▼ 1	grok-3-mini-beta闭源xAI	1,310	±7	5,445	—
33	—	gemini-1.5-pro-001闭源Google	1,310	±2	82,435	—
34	—	gpt-4.1-nano-2025-04-14闭源OpenAI	1,308	±7	6,302	—
35	—	claude-3-opus-20240229闭源Anthropic	1,308	±2	202,641	—
36	—	gpt-4-1106-preview闭源OpenAI	1,303	±2	103,748	—
37	▲ 1	llama-3.3-70b-instruct开源Meta	1,303	±3	43,904	—
38	▼ 1	gpt-4o-mini-2024-07-18闭源OpenAI	1,303	±2	72,536	—
39	—	claude-3-5-haiku-20241022闭源Anthropic	1,303	±3	43,441	—
40	—	gpt-4-0125-preview闭源OpenAI	1,300	±2	97,079	—
41	—	gemini-1.5-flash-002闭源Google	1,298	±3	37,021	—
42	—	gemma-3n-e4b-it开源Google	1,298	±8	3,913	—
43	—	grok-2-mini-2024-08-13闭源xAI	1,293	±2	55,442	—
44	—	gemma-3-4b-it开源Google	1,289	±7	4,321	—
45	—	llama-3.1-nemotron-70b-instruct开源NVIDIA	1,285	±6	7,577	—
46	—	llama-3.1-70b-instruct开源Meta	1,283	±2	58,637	—
47	—	llama-3.1-tulu-3-70b开源Ai2	1,278	±9	3,010	—
48	—	amazon-nova-pro-v1.0闭源Amazon	1,277	±4	26,371	—
49	—	gemma-2-27b-it开源Google	1,276	±2	79,538	—
50	—	gemini-1.5-flash-001闭源Google	1,273	±2	65,661	—
51	—	llama-3.1-nemotron-51b-instruct开源NVIDIA	1,273	±8	3,889	—
52	—	gpt-4-0314闭源OpenAI	1,268	±3	55,962	—
53	—	claude-3-sonnet-20240229闭源Anthropic	1,266	±2	113,067	—
54	—	nemotron-4-340b-instruct开源NVIDIA	1,265	±3	20,608	—
55	—	llama-3-70b-instruct开源Meta	1,264	±2	163,629	—
56	—	gpt-4-0613闭源OpenAI	1,254	±2	91,614	—
57	—	amazon-nova-lite-v1.0闭源Amazon	1,250	±3	20,646	—
58	—	gemma-2-9b-it开源Google	1,249	±3	57,197	—
59	—	gemini-1.5-flash-8b-001闭源Google	1,248	±3	37,697	—
60	—	claude-3-haiku-20240307闭源Anthropic	1,247	±2	122,309	—
61	—	phi-4开源Microsoft	1,246	±3	25,213	—
62	—	olmo-2-0325-32b-instruct开源Ai2	1,243	±9	3,460	—
63	—	amazon-nova-micro-v1.0闭源Amazon	1,232	±3	20,654	—
64	▲ 1	gemini-pro-dev-api闭源Google	1,216	±4	18,800	—
65	▼ 1	llama-3.1-tulu-3-8b开源Ai2	1,216	±10	3,074	—
66	—	llama-3-8b-instruct开源Meta	1,211	±2	109,056	—
67	—	gpt-3.5-turbo-0125闭源OpenAI · 另有 1 个档位	1,206	±3	68,867	—
68	—	granite-3.1-8b-instruct开源IBM	1,204	±9	3,289	—
69	—	llama-3.1-8b-instruct开源Meta	1,202	±2	52,578	—
70	—	gemini-pro闭源Google	1,200	±6	6,557	—
71	—	phi-3-medium-4k-instruct开源Microsoft	1,187	±3	26,105	—
72	—	gemma-2-2b-it开源Google	1,187	±3	48,892	—
73	—	dbrx-instruct-preview开源Databricks	1,182	±4	33,743	—
74	—	granite-3.0-8b-instruct开源IBM	1,178	±7	7,002	—
75	—	wizardlm-70b开源Microsoft	1,174	±6	8,383	—
76	—	granite-3.1-2b-instruct开源IBM	1,172	±9	3,380	—
77	—	snowflake-arctic-instruct开源Snowflake	1,171	±3	34,173	—
78	—	gemma-1.1-7b-it开源Google	1,168	±4	25,070	—
79	—	phi-3-small-8k-instruct开源Microsoft	1,167	±4	18,476	—
80	—	llama-3.2-3b-instruct开源Meta	1,162	±7	8,390	—
81	—	llama-2-70b-chat开源Meta	1,151	±3	39,595	—
82	—	granite-3.0-2b-instruct开源IBM	1,150	±8	7,191	—
83	—	llama2-70b-steerlm-chat开源NVIDIA	1,142	±9	3,636	—
84	—	phi-3-mini-4k-instruct-june-2024开源Microsoft	1,137	±5	12,808	—
85	—	wizardlm-13b开源Microsoft	1,132	±7	7,176	—
86	—	phi-3-mini-4k-instruct开源Microsoft	1,125	±4	21,097	—
87	—	llama-2-13b-chat开源Meta	1,123	±5	19,722	—
88	—	phi-3-mini-128k-instruct开源Microsoft	1,117	±4	21,622	—
89	—	gemma-7b-it开源Google	1,117	±6	9,176	—
90	—	codellama-34b-instruct开源Meta	1,116	±7	7,509	—
91	—	llama-3.2-1b-instruct开源Meta	1,114	±7	8,523	—
92	—	palm-2闭源Google	1,111	±8	8,713	—
93	—	gemma-1.1-2b-it开源Google	1,109	±6	11,351	—
94	—	codellama-70b-instruct开源Meta	1,107	±18	1,192	—
95	—	llama-2-7b-chat开源Meta	1,094	±4	14,532	—
96	—	gemma-2b-it开源Google	1,077	±8	4,918	—
97	—	olmo-7b-instruct开源Ai2	1,063	±8	6,503	—
98	—	dolly-v2-12b开源Databricks	950	±10	3,480	—
99	—	llama-13b开源Meta	935	±12	2,446	—

其余 49 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