Arena Vision
- What does it measure?
- A ranking of which AI gives the better answer when shown a photo, chart or screenshot and asked about it. A high score means the AI is strong at reading and explaining what is in an image.
- Who evaluates it, and how?
- In the Vision section of Arena Intelligence, a user uploads an image and asks a question, two anonymous AIs answer, and the user votes for the better answer. Votes are scored with the same statistical model as text Arena, and only AIs that accept image input take part.
- Reading the score
- This is a head-to-head record, not an exam score. It is computed from the probability of being preferred over other AIs, so there is no maximum; a 100-point gap means the leader is preferred about 64% of the time. Gaps of a few dozen points are effectively the same tier even if the rank differs. Values in this table are currently about 1,100–1,350. Answers are not graded for correctness, so check separately how accurately an AI reads numbers from charts.
Rank
Top score1325
Models tested66
Scoring basisHigher is betterImage-input models only
Last updated2026.11.19
Model release date
Can it spot the wrong step from assembly photos?Furniture Assembly
Claude Fable 5.2🥇
Claude Fable 5.2🥇Furniture Assembly
Claude Fable 5.2🥇Can it draw a floor plan from a few photos of a home?Blueprint-Bench 2
GPT-6.2 Astra🥇
GPT-6.2 Astra🥇Blueprint-Bench 2
GPT-6.2 Astra🥇
Which AI gives the photo and chart answers people prefer most?
Arena Vision
- 🥇AnthropicClaude Fable 5132535.8%38.6%
- 🥈AnthropicClaude Fable 5.1132270.0%41.9%
- 🥉AnthropicClaude Opus 5.5132283.3%51.2%
- #4AnthropicClaude Opus 4.7131733.3%24.5%
- #5GoogleGemini 3.8 Flash131631.7%38.6%
- #6GoogleGemini 3.7 Flash131426.7%—
- #7AnthropicClaude Fable 5.2131390.0%56.6%
- #8AnthropicClaude Opus 4.6131228.3%—
- #9GoogleGemini 3.5 Flash1306—33.6%
- #9MetaMuse Spark 1.31306——
- #11MetaMuse Spark1306——
- #12MetaMuse Spark 1.21304——
- #13OpenAIGPT-5.4130137.5%27.1%
- #14GoogleGemini 3.6 Flash129723.3%31.2%
- #15GoogleGemini 3.1 Pro129726.7%26.5%
- #15Z.aiGLM 5.3 Flash1297——
- #17OpenAIGPT-5.5129644.2%36.2%
- #18MetaMuse Spark 1.11291——
- #19AnthropicClaude Opus 4.8128842.5%14.5%
- #20SpaceXSIGrok 4.5128822.5%27.3%
- #21AnthropicClaude Sonnet 5.5128675.0%39.3%
- #22GoogleGemini 3 Flash1285—0.0%
- #23OpenAIGPT-6.1 Sol128480.0%43.2%
- #24AnthropicClaude Sonnet 4.61283—6.7%
- #25Moonshot AIKimi K2.6128221.7%3.9%
- #26OpenAIGPT-5.6 Sol128256.7%33.6%
- #27GoogleGemma 4 31B1279——
- #28DeepSeekDeepSeek V4.1 Flash127834.2%—
- #29AlibabaQwen3.7 Plus1277——
- #30OpenAIGPT-6 Astra127680.0%49.7%
- #31ByteDanceDola Seed 2.0 Pro1274——
- #32AnthropicClaude Sonnet 51272—24.9%
- #33OpenAIGPT-5.6 Terra127054.2%30.8%
- #34Moonshot AIKimi K2.51269——
- #35GoogleGemini 3.5 Flash-Lite1268——
- #36XiaomiMiMo-V2.6-Pro1266——
- #37SpaceXSIGrok 4.6126540.0%33.2%
- #38StepfunStep-51265——
- #39GoogleGemini 2.5 Pro1263——
- #40SpaceXSIGrok 4.20 (Reasoning)1263—0.0%
- #41AlibabaQwen3.5 397B A17B1262——
- #42XiaomiMiMo-V2.6-Flash1258——
- #43OpenAIGPT-5.6 Luna125642.5%22.6%
- #44MiniMaxMiniMax M31248——
- #45XiaomiMiMo V2.51247——
- #46OpenAIGPT-6.1 Terra124779.2%42.1%
- #47Mistral AIMistral Large 4124630.8%—
- #48OpenAIGPT-5.4 Mini1245——
- #49GoogleGemini 3.1 Flash Lite1240——
- #50GoogleGemini 2.5 Flash1235——
- #51OpenAIGPT-6 Sol123258.3%36.9%
- #52SpaceXSIGrok 4.7123120.8%32.5%
- #53SpaceXSIGrok 4.31228—0.0%
- #54Mistral AIMistral Medium 3.51222——
- #55OpenAIGPT-4.11211——
- #56OpenAIGPT-6 Luna121044.2%31.2%
- #57OpenAIGPT-51208——
- #58OpenAIGPT-5 Mini1203——
- #59SpaceXSIGrok 4.1 Fast (Reasoning)1202——
- #60GoogleGemini 2.5 Flash Lite1198——
- #61OpenAIGPT-5.4 Nano1195——
- #62AnthropicClaude Opus 41192——
- #63AnthropicClaude Sonnet 41191——
- #64OpenAIGPT-5 Nano1159——
- #65MetaLlama 4 Maverick1142——
- #66MetaLlama 4 Scout1119——
Half of models ≤ 1270Measured by: User preference votes
Scores from an earlier edition
These scores are not in the latest leaderboard (2026.11.19) — the model was dropped by the source or has not been refreshed yet. The last score we received is shown for reference, unranked.
- as of 2026.11.09$3.301266
37 of the 46 AIs released in the last 60 days have no score here yet · Gemini 4 Neon, GPT-6.2 Astra, Microsoft-Decision-1.1 and 34 more
It’s a preference vote, not a correctness check. A wrong but friendly answer can win.
Which questions got asked depends on who voted — they may differ from your use.
Scores labelled 'Self-reported' were published by the AI maker itself, not measured by a third party.