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

A screenshot of a new benchmark claims Gemini 4 Argon leads the rankings, though the post gives no scores or test details.

A leak claims the Gemini 3.8 Flash endpoint may actually run Gemini 4 Argon, meaning users could already be using the next-gen model without knowing.

He puts a 55% probability on Gemini Argon beating both Fable and Astra while costing 5x less, which he says would force both to cut prices.

A user asks whether Gemini 4 Argon or Fable 5.5 ships first and predicts the order will be Fable 5.5, GPT-6.1 Astra, then Gemini 4 Flash.

A poll asks whether people expect Gemini 4 Argon or Gemini 4 Carbon to be the better model, or whether both will flop.

A leak says Gemini 4 Carbon has appeared on Google's internal coding platform Jetski, with the internal codename sequence running Argon, Barium, Carbon, while Argon is still not public.

A poll post asks which model people would choose if usage were unlimited among Claude Opus 5.5, GPT-6 Astra, Gemini 4 Pro and Grok, with no hands-on testing.

A user complains that Gemini ignores instructions not to use web search and hopes Argon and upcoming Gemini web updates will let people turn off search and other tools manually.

A meme video jokes about what happens when you say "Claude is better" in a comment section full of Codex and Gemini Argon fans.

Before Gemini 4 Argon has reached everyone, a new codename, Carbon, is reportedly circulating inside Google, fueling speculation about unreleased versions.

Google AI Studio lead Logan Kilpatrick reportedly confirmed the release order: Gemini 3.5 Pro first, then the flagship Gemini 4 Argon.

A user says Gemini 4 Argon mapped a real Xbox controller and let them interact with it using their own controller, calling the details surprisingly good.