Particle.news
Download on the App Store
3 ARTICLES
·
3h ago

Google Debuts Gemini 2.5 Computer Use for Developers as It Formalizes Tiered Usage Caps

The developer-only release underscores a cautious, web-first approach with limited actions and sped-up demos.

Image
Gemini kann wohl bald auch Browser selbstständig steuern. (Bild: Shutterstock/DIA TV)

Overview

  • Gemini 2.5 Computer Use is now accessible through Google AI Studio and Vertex AI, with a public hands-on demo available via Browserbase.
  • The agent executes tasks by analyzing screenshots and issuing function calls to click, type, drag, and request confirmations, iterating until completion.
  • Google notes current constraints: about 13 supported actions, performance tuned for web interfaces over mobile, and no full PC control, with demo videos accelerated 3×.
  • Google’s Help Center now lists concrete limits, including free accounts capped at 5 prompts per day on Gemini 2.5 Pro, a 32,000‑token context window, and up to 100 image generations or edits daily.
  • Paid plans raise ceilings to 100 prompts/day (Pro) and 500 (Ultra) with 1‑million‑token windows, up to 1,000 images/day, Pro/Ultra video quotas, Deep Research caps (5/month free; 20/day Pro; 200/day Ultra), and Ultra‑only Deep Think at 10 prompts/day with a 192,000‑token window.