Vision Arena: VLM Leaderboard & Best Vision Models

Vision Arena🏆Overall

View overall rankings across multimodal AI models capable of reasoning over visual inputs.

Oct 9, 2026

1,462,389 votes

49 open source models

				Rank Spread
1	35	glm-5.3-flash Z.ai · MIT	1273±7	1147	11,157	$0.06 / $0.20	1M
2	41	kimi-k2.6 Moonshot · Modified MIT	1265±7	3255	16,379	$0.95 / $4	262.1K
3	44	gemma-4-31b Google · Apache 2.0	1262±6	3458	39,669	$0.99 / $1.49	131.1K
4	46	deepseek-v4.1-flash-max DeepSeek · MIT	1260±11	3362	3,559	$0.30 / $1.20	1M
5	51	kimi-k2.5-thinking Moonshot · Modified MIT	1252±6	4167	28,297	$0.60 / $3	N/A
6	56	mimo-v2.6-flash Xiaomi · MIT	1248±12	3874	2,908	$0.14 / $0.28	1.1M
7	59	mimo-v2.6-pro Xiaomi · MIT	1246±15	3777	1,778	$0.43 / $0.87	1.1M
8	60	qwen3.5-397b-a17b Alibaba · Apache 2.0	1246±5	4769	32,058	$0.50 / $3.60	262.1K
9	61	qwen3.8-27b Alibaba · Apache 2.0	1244±8	4674	7,013	$0.40 / $3	262.1K
10	64	gemma-4-26b-a4b Google · Apache 2.0	1242±6	5174	23,943	N/A	N/A
11	68	kimi-k2.5-instant Moonshot · Modified MIT	1236±11	5178	4,114	$0.50 / $2.50	262.1K
12	69	minimax-m3 MiniMax · MiniMax Community License	1236±6	5678	19,026	$0.30 / $1.20	524.3K
13	71	mimo-v2.5 Xiaomi · MIT	1234±6	5778	28,794	$0.14 / $0.28	1.1M
14	75	qwen3.5-122b-a10b Alibaba · Apache 2.0	1226±7	6486	13,241	$0.26 / $2.08	262.1K
15	79	qwen3.5-27b Alibaba · Apache 2.0	1217±6	7488	29,644	$0.26 / $2.60	262.1K
16	83	qwen3-vl-235b-a22b-instruct Alibaba · Apache 2.0	1215±7	7491	12,751	$0.29 / $1.15	262.1K
17	86	Inkling Small Thinky · Apache 2.0	1208±8	7799	7,472	$0.45 / $1.20	524.3K
18	91	mistral-medium-3.5 Mistral · Modified MIT	1200±10	80103	5,375	$0.75 / $3.75	262.1K
19	92	mistral-large-3 Mistral · Apache 2.0	1200±8	84101	8,643	$0.50 / $1.50	N/A
20	97	qwen3-vl-235b-a22b-thinking Alibaba · Apache 2.0	1189±12	86107	2,386	$0.40 / $4	131.1K
21	109	glm-4.6v Z.ai · MIT	1161±14	102123	2,457	$0.30 / $0.90	131.1K
22	111	gemma-3-27b-it Google · Gemma	1160±8	105122	16,985	$0.08 / $0.45	131.1K
23	116	glm-4.5v Z.ai · MIT	1153±12	107125	3,249	$0.60 / $1.80	65.5K
24	119	llama-4-maverick-17b-128e-instruct Meta · Llama 4	1145±9	108127	6,922	$0.63 / $1.80	131.1K
25	120	step-3 StepFun · Apache 2.0	1145±12	108129	3,196	$0.57 / $1.42	65.5K
26	123	mistral-small-2506 Mistral · Apache 2.0	1140±9	112132	10,846	$0.10 / $0.30	262.1K
27	125	llama-4-scout-17b-16e-instruct Meta · Llama	1127±10	117134	6,451	$0.05 / $0.10	16.4K
28	126	mistral-small-3.1-24b-instruct-2503 Mistral · Apache 2.0	1127±8	118134	28,493	$0.10 / $0.30	32K
29	129	qwen2.5-vl-72b-instruct Alibaba · Qwen	1121±11	122134	3,768	$0.80 / $1	128K
30	131	qwen2.5-vl-32b-instruct Alibaba · Apache 2.0	1119±15	120137	1,490	$0.20 / $0.60	128K
31	133	molmo-2-8b Ai2 · Apache 2.0	1112±20	122139	1,189	$0.20 / $0.20	36.9K
32	136	pixtral-large-2411 Mistral · MRL	1095±9	132139	5,423	$2 / $6	128K
33	139	qwen2-vl-72b Alibaba · Qwen	1084±10	134141	5,937	$0.90 / $0.90	32.8K
34	144	molmo-72b-0924 Ai2 · Apache 2.0	1045±13	141151	3,048	N/A	N/A
35	146	llama-3.2-vision-90b-instruct Meta · Llama 3.2	1032±9	144152	8,682	N/A	N/A
36	147	qwen2-vl-7b-instruct Alibaba · Apache 2.0	1031±11	144154	5,766	$0.20 / $0.20	32.8K
37	148	pixtral-12b-2409 Mistral · Apache 2.0	1025±10	144154	7,511	$0.15 / $0.15	128K
38	149	internvl2-26b MIT	1025±13	144155	5,148	N/A	N/A
39	154	c4ai-aya-vision-32b Cohere · CC-BY-NC-4.0	1000±22	146161	847	N/A	N/A
40	156	molmo-7b-d-0924 Ai2 · Apache 2.0	993±14	150159	2,815	N/A	N/A
41	157	llama-3.2-vision-11b-instruct Meta · Llama 3.2	993±11	152159	4,817	N/A	N/A
42	158	nvila-internal-15b-v1 Nvidia · -	985±20	150163	1,077	N/A	N/A
43	159	llava-onevision-qwen2-72b-ov Apache 2.0	978±18	153163	1,321	N/A	N/A
44	160	llava-v1.6-34b Apache 2.0	964±12	158163	4,531	N/A	N/A
45	161	cogvlm2-llama3-chat-19b Z.ai · CogVLM2	963±15	157163	1,991	N/A	N/A
46	162	minicpm-v-2_6 Apache 2.0	962±16	157163	1,987	N/A	N/A
47	163	internvl2-4b MIT	958±13	158163	3,703	N/A	N/A
48	164	phi-3.5-vision-instruct Microsoft · MIT	919±16	164164	2,592	$0.13 / $0.52	128K
49	165	phi-3-vision-128k-instruct Microsoft · MIT	881±19	165165	1,401	$0.20 / $0.20	32.1K

Default Leaderboard Plots

1 model with no battles against the other models is hidden.

Vision Arena🏆Overall

View overall rankings across multimodal AI models capable of reasoning over visual inputs.

Default Leaderboard Plots

Fraction of Model A Wins in A vs. B Battles (Ties Excluded)

Average Win Rate Against Each Opponent (Ties Excluded)

Battle Count for Each Pair of Models (Ties Excluded)

Confidence Intervals on Model Strength (via Bootstrapping)