Vision Model Leaderboard: Best Vision-Language Models

Ranked by LMArena Vision Arena scores: users upload an image, ask a question, and vote between two anonymous models’ answers.

Overall OCR Diagrams Homework Captioning Entity recognition Creative writing Humor English Chinese
#1 EnglishData Oct 19, 2026
Claude Opus 4.6 (High)Anthropic · 1324

In LMArena Vision Arena data published Oct 19, 2026, Anthropic’s Claude Fable 5 (High) ranks #1 of 167 models with 1325 (14,714 votes).

Full leaderboard · English

50 models
All Open weights

Sorted by LMArena score by default; ± is half the 95% confidence interval. Category columns show the model’s score on that sub-board; – means no score.

# Model LMArena · English License
Score Rank Votes
1 GLM-5.3-FlashZ.aiOpen 1295±10 18 5,043 MIT
2 DeepSeek-V4.1-Flash-20260910 (Max)DeepSeekOpen 1288±17 32 1,301 MIT
3 Kimi K2.6Moonshot AIOpen 1287±9 34 6,756 Modified MIT
4 Kimi K2.5 (Thinking)Moonshot AIOpen 1274±8 44 10,972 Modified MIT
5 Gemma 4 31BGoogleOpen 1271±8 47 15,807 Apache 2.0
6 MiMo-V2.6-FlashXiaomiOpen 1269±18 49 1,134 MIT
7 Qwen3.5-397B-A17BAlibabaOpen 1268±7 51 12,659 Apache 2.0
8 MiMo-V2.6-ProXiaomiOpen 1268±24 52 612 MIT
9 Qwen3.8-27BAlibabaOpen 1267±13 53 2,445 Apache 2.0
10 MiniMax-M3MiniMaxOpen 1263±9 56 7,372 MiniMax Community License
11 MiMo-V2.5XiaomiOpen 1259±8 60 11,479 MIT
12 Kimi K2.5 (Instant)Moonshot AIOpen 1258±16 62 1,419 Modified MIT
13 Gemma 4 26B A4BGoogleOpen 1256±9 63 9,635 Apache 2.0
14 Qwen3-VL-235B-A22B-InstructAlibabaOpen 1256±10 64 4,605 Apache 2.0
15 Qwen3.5-122B-A10BAlibabaOpen 1249±10 68 5,296 Apache 2.0
16 Qwen3.5-27BAlibabaOpen 1246±8 69 11,800 Apache 2.0
17 TInkling-SmallThinking MachinesOpen 1243±12 73 2,816 Apache 2.0
18 Mistral Medium 3.5Mistral AIOpen 1240±14 75 2,314 Modified MIT
19 NNVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)NVIDIAOpen 1239±21 77 864 OpenMDW-1.1
20 Mistral Large 3Mistral AIOpen 1235±12 79 3,184 Apache 2.0
21 Qwen3-VL-235B-A22B-ThinkingAlibabaOpen 1215±18 92 972 Apache 2.0
22 SStep-3StepFunOpen 1199±15 99 1,616 Apache 2.0
23 Gemma 3 27B ITGoogleOpen 1188±10 105 7,938 Gemma
24 GLM-4.5VZ.aiOpen 1183±16 108 1,596 MIT
25 GLM-4.6VZ.aiOpen 1179±20 111 914 MIT
26 Mistral Small 3.2 (2506)Mistral AIOpen 1162±11 117 5,376 Apache 2.0
27 Qwen2.5-VL-32B-InstructAlibabaOpen 1158±19 119 870 Apache 2.0
28 Mistral Small 3.1 24B Instruct (2503)Mistral AIOpen 1157±10 121 14,751 Apache 2.0
29 Llama 4 Maverick 17B 128E InstructMetaOpen 1150±12 123 3,387 Llama 4
30 Llama 4 Scout 17B 16E InstructMetaOpen 1143±12 126 3,098 Llama
31 Qwen2.5-VL-72B-InstructAlibabaOpen 1115±14 131 2,169 Qwen
32 Pixtral Large (2411)Mistral AIOpen 1100±12 133 2,921 MRL
33 Molmo 2 8BAi2Open 1092±29 135 461 Apache 2.0
34 Qwen2-VL-72BAlibabaOpen 1054±13 143 3,230 Qwen
35 Molmo 72B (0924)Ai2Open 1045±17 144 1,627 Apache 2.0
36 OInternVL2 26BOpenGVLabOpen 1037±15 145 3,250 MIT
37 Pixtral 12B (2409)Mistral AIOpen 1027±12 147 4,090 Apache 2.0
38 Llama 3.2 90B Vision InstructMetaOpen 1013±12 150 4,630 Llama 3.2
39 Molmo 7B-D (0924)Ai2Open 1013±18 151 1,530 Apache 2.0
40 Qwen2-VL-7B-InstructAlibabaOpen 1006±14 153 3,028 Apache 2.0
41 Aya Vision 32BCohereOpen 999±30 155 447 CC-BY-NC-4.0
42 Llama 3.2 11B Vision InstructMetaOpen 997±15 156 2,575 Llama 3.2
43 OInternVL2 4BOpenGVLabOpen 986±16 158 2,189 MIT
44 LLLaVA 1.6 34BOpen 970±16 160 2,813 Apache 2.0
45 CogVLM2 Llama3 Chat 19BZ.aiOpen 969±20 161 1,162 CogVLM2
46 LLLaVA-OneVision Qwen2 72B OVOpen 962±22 162 841 Apache 2.0
47 Nnvila-internal-15b-v1NVIDIAOpen 959±26 163 619 -
48 OMiniCPM-V 2.6OpenBMBOpen 949±19 165 1,256 Apache 2.0
49 MPhi-3.5 Vision InstructMicrosoftOpen 875±19 166 1,608 MIT
50 MPhi-3 Vision 128K InstructMicrosoftOpen 837±23 167 880 MIT

Scores: LMArena leaderboard-dataset (lmarena-ai on Hugging Face, CC-BY-4.0, vision config), English published Oct 19, 2026. Please credit the source when citing.

Sources & methodology

  1. Scores are Bradley-Terry Arena ratings fitted from human votes between two anonymous answers to the same image prompt.
  2. Categories (OCR, diagrams, homework, captioning, Chinese/English…) are separate LMArena boards computed on prompts of that type, not a breakdown of the overall score; ranks can differ a lot.
  3. Fields without public data (price, context length, release date…) are not shown; blanks are –. Nothing is estimated.
  4. Synced daily. JSON API: /api/public/vision-leaderboard.php

This week

Collecting data: 1/7 days recorded; weekly changes appear after 7 days

Synced automatically once a day; ranks are compared only between real snapshots, never backfilled.

FAQ

What is the best vision-language model right now?

In LMArena Vision Arena data published Oct 19, 2026, Anthropic’s Claude Fable 5 (High) ranks #1 of 167 models with 1325 (14,714 votes).

What is each vendor’s best vision model?

Top overall entry per vendor: Anthropic Claude Fable 5 (High) (1325, #1), Google Gemini 3.8 Flash (High) (1315, #7), Alibaba Qwen3.8-Max (1313, #10), Meta Muse Spark 1.3 (Max) (1308, #12), OpenAI GPT-5.4 (High) (1303, #18).

Which model is best at OCR?

On LMArena’s OCR board (123 models), Claude Fable 5 (High) is #1 with 1337.

What is the best open-weight vision model?

50 open-weight models are listed; the top one is GLM-5.3-Flash (1294, MIT).

How is this different from benchmarks like MMMU?

MMMU and similar benchmarks auto-score a fixed question set; this board comes from blind human votes on people’s own images and questions, reflecting real-use preference. They are not convertible.