Vision Arena

View rankings across multimodal, generative AI models capable of understanding and processing visual inputs

Last Updated

Jan 9, 2026

Total Votes

586,936

Total Models

/

Style Control

	Rank Spread
1	1◄─►1	gemini-3-pro	1302	±11	5,564	Google	Proprietary
2	2◄─►3	gemini-3-flash	1274	±13	2,630	Google	Proprietary
3	2◄─►6	gemini-3-flash (thinking-minimal)	1264	±20	949	Google	Proprietary
4	3◄─►6	gemini-2.5-pro	1249	±7	66,716	Google	Proprietary
5	3◄─►9	gpt-5.1-high	1247	±12	3,291	OpenAI	Proprietary
6	3◄─►11	gpt-5.1	1239	±12	3,510	OpenAI	Proprietary
7	5◄─►11	chatgpt-4o-latest-20250326	1235	±7	18,980	OpenAI	Proprietary
8	5◄─►19	gpt-4.5-preview-2025-02-27	1225	±11	2,925	OpenAI	Proprietary
9	6◄─►19	gpt-5-chat	1223	±8	38,888	OpenAI	Proprietary
10	6◄─►19	gemini-2.5-flash-preview-09-2025	1223	±10	4,993	Google	Proprietary
11	5◄─►22	ernie-5.0-preview-1220	1222	±15	1,816	Baidu	Proprietary
12	8◄─►19	o3-2025-04-16	1217	±7	45,237	OpenAI	Proprietary
13	8◄─►22	gpt-4.1-2025-04-14	1215	±7	40,349	OpenAI	Proprietary
14	8◄─►22	gemini-2.5-flash	1213	±7	41,507	Google	Proprietary
15	8◄─►25	qwen3-vl-235b-a22b-instruct	1211	±9	6,635	Alibaba	Apache 2.0
16	8◄─►25	gpt-5-high	1209	±8	33,915	OpenAI	Proprietary
17	8◄─►29	claude-opus-4-20250514-thinking-16k	1209	±15	1,416	Anthropic	Proprietary
18	8◄─►29	ernie-5.0-preview-1120	1205	±11	3,958	Baidu	Proprietary
19	8◄─►30	claude-sonnet-4-20250514-thinking-32k	1204	±16	1,307	Anthropic	Proprietary
20	12◄─►29	gpt-4.1-mini-2025-04-14	1201	±8	39,472	OpenAI	Proprietary
21	12◄─►29	o4-mini-2025-04-16	1201	±8	39,967	OpenAI	Proprietary
22	12◄─►34	claude-3-7-sonnet-20250219-thinking-32k	1194	±15	1,626	Anthropic	Proprietary
23	15◄─►34	hunyuan-vision-1.5-thinking	1193	±12	2,731	Tencent	Proprietary
24	15◄─►34	o1-2024-12-17	1192	±10	3,694	OpenAI	Proprietary
25	15◄─►34	qwen3-vl-235b-a22b-thinking	1190	±12	2,523	Alibaba	Apache 2.0
26	17◄─►34	gemini-2.5-flash-lite-preview-06-17-thinking	1188	±8	36,702	Google	Proprietary
27	17◄─►34	claude-opus-4-20250514	1188	±12	2,511	Anthropic	Proprietary
28	17◄─►35	claude-sonnet-4-20250514	1186	±13	2,031	Anthropic	Proprietary
29	21◄─►35	gpt-5-mini-high	1184	±9	28,880	OpenAI	Proprietary
30	17◄─►35	qwen-vl-max-2025-08-13	1183	±12	3,314	Alibaba	Proprietary
31	22◄─►35	grok-4-0709	1180	±8	31,497	xAI	Proprietary
32	22◄─►35	gemini-1.5-pro-002	1177	±8	8,916	Google	Proprietary
33	22◄─►39	gemini-2.5-flash-lite-preview-09-2025-no-thinking	1177	±10	5,089	Google	Proprietary
34	22◄─►39	claude-3-7-sonnet-20250219	1176	±9	4,659	Anthropic	Proprietary
35	28◄─►42	gemini-2.0-flash-001	1168	±7	9,733	Google	Proprietary
36	33◄─►46	gpt-4o-2024-05-13	1161	±8	23,280	OpenAI	Proprietary
37	33◄─►46	claude-3-5-sonnet-20241022	1161	±7	10,527	Anthropic	Proprietary
38	33◄─►50	glm-4.5v	1157	±12	3,474	Z.ai	MIT
39	35◄─►50	mistral-medium-2505	1155	±9	11,421	Mistral	Proprietary
40	35◄─►50	gemma-3-27b-it	1155	±8	18,261	Google	Gemma
41	33◄─►51	step-1o-turbo-202506	1154	±14	2,052	StepFun	Proprietary
42	35◄─►54	hunyuan-large-vision	1148	±16	1,459	Tencent	Proprietary
43	36◄─►51	mistral-medium-2508	1147	±8	34,754	Mistral	Proprietary
44	36◄─►51	claude-3-5-sonnet-20240620	1146	±9	21,632	Anthropic	Proprietary
45	36◄─►53	gpt-5-nano-high	1146	±12	4,097	OpenAI	Proprietary
46	36◄─►51	llama-4-maverick-17b-128e-instruct	1146	±9	7,157	Meta	Llama 4
47	38◄─►57	step-3	1140	±12	3,438	StepFun	Apache 2.0
48	38◄─►57	mistral-small-2506	1139	±9	11,633	Mistral	Apache 2.0
49	38◄─►57	gemini-1.5-flash-002	1138	±9	7,256	Google	Proprietary
50	38◄─►59	claude-3-5-haiku-20241022	1134	±15	1,564	Anthropic	Proprietary
51	41◄─►59	gemini-2.0-flash-lite-preview-02-05	1133	±10	4,001	Google	Proprietary
52	45◄─►59	llama-4-scout-17b-16e-instruct	1125	±10	6,623	Meta	Llama
53	46◄─►59	mistral-small-3.1-24b-instruct-2503	1125	±9	27,468	Mistral	Apache 2.0
54	45◄─►59	step-1o-vision-32k-highres	1123	±12	2,844	StepFun	Proprietary
55	47◄─►59	qwen2.5-vl-72b-instruct	1121	±10	3,767	Alibaba	Qwen
56	47◄─►60	gpt-4o-2024-08-06	1118	±12	3,384	OpenAI	Proprietary
57	50◄─►60	gemini-1.5-pro-001	1117	±11	16,735	Google	Proprietary
58	47◄─►62	qwen2.5-vl-32b-instruct	1116	±15	1,490	Alibaba	Apache 2.0
59	50◄─►62	gpt-4-turbo-2024-04-09	1112	±11	13,390	OpenAI	Proprietary
60	58◄─►64	gpt-4o-mini-2024-07-18	1097	±7	17,362	OpenAI	Proprietary
61	58◄─►64	pixtral-large-2411	1093	±9	5,439	Mistral	MRL
62	56◄─►68	gpt-4.1-nano-2025-04-14	1088	±18	1,211	OpenAI	Proprietary
63	60◄─►66	qwen2-vl-72b	1085	±9	5,946	Alibaba	Qwen
64	60◄─►68	qwen-vl-max-1119	1084	±16	1,431	Alibaba	Proprietary
65	62◄─►69	gemini-1.5-flash-8b-001	1070	±10	6,255	Google	Proprietary
66	63◄─►70	claude-3-opus-20240229	1064	±10	15,568	Anthropic	Proprietary
67	62◄─►70	step-1v-32k	1061	±16	1,551	StepFun	Proprietary
68	63◄─►70	gemini-1.5-flash-001	1059	±11	13,261	Google	Proprietary
69	66◄─►73	molmo-72b-0924	1047	±13	3,063	AI2	Apache 2.0
70	65◄─►77	hunyuan-standard-vision-2024-12-31	1043	±21	809	Tencent	Proprietary
71	69◄─►77	llama-3.2-vision-90b-instruct	1032	±8	8,708	Meta	Llama 3.2
72	69◄─►79	qwen2-vl-7b-instruct	1031	±10	5,778	Aliaba	Apache 2.0
73	70◄─►79	pixtral-12b-2409	1024	±9	7,520	Mistral	Apache 2.0
74	69◄─►80	internvl2-26b	1024	±12	5,146	OpenGVLab	MIT
75	70◄─►82	amazon-nova-lite-v1.0	1020	±15	1,856	Amazon	Proprietary
76	70◄─►82	amazon-nova-pro-v1.0	1020	±13	2,338	Amazon	Proprietary
77	70◄─►81	claude-3-sonnet-20240229	1019	±11	12,314	Anthropic	Proprietary
78	72◄─►84	yi-vision	1004	±18	1,220	01 AI	Proprietary
79	74◄─►84	claude-3-haiku-20240307	1002	±12	13,380	Anthropic	Proprietary
80	72◄─►86	c4ai-aya-vision-32b	1000	±22	847	Cohere	CC-BY-NC-4.0
81	75◄─►84	molmo-7b-d-0924	996	±13	2,828	AI2	Apache 2.0
82	78◄─►84	llama-3.2-vision-11b-instruct	991	±11	4,818	Meta	Llama 3.2
83	76◄─►88	nvila-internal-15b-v1	987	±20	1,078	NVIDIA	-
84	78◄─►88	llava-onevision-qwen2-72b-ov	980	±18	1,321	LLaVA	Apache 2.0
85	83◄─►88	llava-v1.6-34b	966	±12	4,531	LLaVA	Apache 2.0
86	82◄─►88	minicpm-v-2_6	964	±15	1,987	OpenBMB	Apache 2.0
87	82◄─►88	cogvlm2-llama3-chat-19b	964	±15	1,991	Zhipu AI	CogVLM2
88	83◄─►88	internvl2-4b	957	±12	3,706	OpenGVLab	MIT
89	89◄─►89	phi-3.5-vision-instruct	921	±15	2,592	Microsoft	MIT
90	90◄─►90	phi-3-vision-128k-instruct	883	±18	1,401	Microsoft	MIT

Vision Arena

Default Leaderboard Plots

Confidence Intervals on Model Strength (via Bootstrapping)

Battle Count for Each Combination of Models (without Ties)

Average Win Rate Against All Other Models (Uniform Sampling and No Ties)

Fraction of Model A Wins for All Non-tied A vs. B Battles