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

An image post showing Gemini 4 Argon's take on the line between ordinary people and elites.

A discussion prompt asking whether Gemini 4 Argon has actually converted Claude users, or whether everyone is still on Claude despite the buzz.

A Japanese user asked Gemini 4 Argon to verify a science video about clearing senescent cells, and it found the content matches real experiments and papers.

A bookmark-style list of 12 AI tools for projects, tagging Gemini as the all-rounder after Argon alongside writing, research, image, video and coding tools.

A user had Gemini 4 Argon and Meta's Muse Wonder build the same game, reporting a surprisingly large gap between the two results.

After DeepSeek's Flash update, Antigravity users want a stronger efficient Flash model, with the poster expecting Gemini 4 Flash to beat current Flash models.

A user says generating one Tesla Roadster video with Gemini Argon 4 used up all their credits, calling it a token sucker.

A creator used Gemini 4 Argon and Flow to auto-produce a one-minute short film, spending 8% of the weekly quota with only one continuity error.

Same prompt across four models: Opus 5.5 took 1h20 and $4.24, Gemini 4 Argon 23 min and $1.50, Muse Wonder 2 min and $0.58, GPT-6.1 Sol 24 min and $0.97.

A user notes Google shipped three Flash models in six weeks and then Gemini 4 Argon about 40 days later, arguing that pace is needed to compete with OpenAI and Anthropic.

A first-time Gemini coder reran a hard task previously given to Claude Opus 5.5 and says Argon handled it better at half the per-token price ($2/$10 vs $4/$20).

Google AI Pro subscribers can now build with Gemini 4 Argon in AI Studio Build and the playground.