Model laboratory
Phone Automation Vision Benchmark
Compare phone-automation models against the evidence they actually saw. Every score resolves to a reviewable image, expected state, and raw response.
In productionFixed in code · every phone task uses it
Loading…
Uploads, warmups, app navigation and the camera check before each task all run on this one model. Runs here compare alternatives; nothing switches models per task.
Overall
—
Coordinates
—
Classification
—
Cost / 1k calls
—
Loading benchmark ledger…