Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaks

Gemini 4 Argon is now usable in Antigravity, but accounts bought cheaply on Xianyu cannot call it, so Pro users get different treatment.

A Portuguese-speaking user says Gemini 4 Argon is a great model, especially for data analysis, and hopes Google won't weaken it in the coming days or weeks.

A Spanish-language post says Gemini 4 Argon is now available in Antigravity to AI Pro users paying $20 per month.

A user reports that Gemini 4 Argon can now be used for coding in Antigravity on the $20 Pro subscription, lowering the barrier to entry.

This week's AI news roundup notes Gemini 4 Argon's official wide release with a 1M token output limit up from 64K, a shared input/output window, and $2/$10 intro pricing until Nov 30.

Pi 1.3.0 updates share destinations, caching and model-switching behavior, and fixes Thinking Level compatibility for Gemini 4 Argon, MiMo, Claude and other models.

Gemini 4 Argon is now on the $20 Gemini plan and usable directly in Antigravity, with a usage limit separate from Gemini 3.8 Flash, Claude and GPT; one user estimates roughly 2–3 hours of continuous use per week.

A user compares Gemini 4 Argon's usage limits with the generous allowances of Opus 5.5, Sol 6.1 and Muse Wonder Full, questioning why Google offers the least despite having the most compute.

A user notes that Gemini 4 Argon usage limits are separate in Antigravity and the Gemini app, and that after hitting the Antigravity cap you can pick Task in a new Gemini app chat and select a local folder to keep working.

A short post says Gemini 4 Argon is now available for Pro users, with what appears to be an interface screenshot.

A user complains that the AI Pro trial bundled with a Pixel purchase may exclude Gemini 4 Argon until the six-month trial ends, and asks @antigravity for help.

Paid Pro users can now use Gemini 4 Argon in Antigravity, and the poster asks how it compares with GPT-6.1 Astra and Opus 5.5 on real coding tasks.