Vibe-Checking Qwen3.8-Max on Hard Visual Grounding Tasks

Everyone's screenshotting Qwen3.8-Max drawing bounding boxes. I used FiftyOne to ground-truth-score it on rotation, outline tracing, and 27,083 real logos instead: the tasks that actually matter if you're pointing this thing at a robot. The FiftyOne App chat panel asking Qwen3.8-Max to point to a Canadian beer cap, with its live thinking trace... Weiterlesen: Vibe-Checking Qwen3.8-Max on Hard Visual Grounding Tasks