Visual Reasoning Benchmark

Each model answers the same reasoning prompts. Answers are graded against the ground truth, and the score is the percentage answered correctly.

Reasoning is run at two levels of reasoning effort. The low row is the model answering with minimal deliberation, the high row is the same model on the same questions allowed to think longer. Both rows are ranked together so you can see whether the extra thinking is worth its cost and latency, which the Est. cost and Speed columns show. For most models it helps, for a few it changes nothing at all, and for a couple the score drops.

Low and high are benchmark tiers mapped to each model's native reasoning mechanism, because no two labs expose the same control. Claude, GPT, and Grok take low or high as a reasoning effort parameter. Gemini runs at its low and high thinking levels (a minimal versus large thinking budget for the older Gemini 2.5 Pro). Qwen, GLM, and Kimi only expose an on/off thinking mode, so low runs them with thinking disabled and high with thinking at high effort. Muse and Qwen3.8-Max cannot switch reasoning off, so their low pass uses the lowest native setting. All six tasks and the Average use the low pass; high has been run only for Reasoning.

Every other table, including the overall Average, uses the low-effort pass. High effort has only been run for Reasoning, so using it elsewhere would compare models on different amounts of compute.

The answer is never written directly in the image. The model has to combine several visual facts, such as reading two values and comparing them, judging spatial relationships, or applying logic to what it sees.

Most models in this leaderboard link to their Playground page. Click the model name to open it, then upload your own image and run it. A few models are benchmarked for comparison only and do not have a Playground page yet.