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

A post with a video tags @GeminiApp to ask when Gemini 4 Pro will launch, highlighting the gap between release pace and user expectations.

A reply gives a personal prediction ranking of Gemini 4 Pro > Opus 5.5 > GPT-6 Astra > Fable 5.1 > Sonnet 5.5, and guesses 4 Flash will be close to Opus 5.5.

Ruby eyes, gears spinning under smoked glass and metallic wing veins, generated in one pass in under 7 minutes, which the author says shows the model understands "make it look expensive".

The author ranks Opus 5.5 above Sonnet 5.5, GPT-6 Astra, Fable 5.1 and Gemini 4 Pro, arguing the competition is too strong and this is already Gemini's best case.

A user tags Google and GeminiApp asking where Gemini 4 Pro is, saying they want a low-priced Gemini coding model.

Replying to Google's official account, the user says they like the current products but are still waiting for Gemini 4 Pro and want it released as soon as possible.

A short post with an image asks "Who did this? Gemini 4 pro......", hinting that a Gemini 4 Pro screenshot or leaderboard is drawing discussion.

The author says Gemini 4 Pro's new checkpoint shows clear gains in design taste and prompt adherence, but was only briefly available between 4 and 5 AM and suspects it has been taken down again.

A video compares outputs from GPT-6.1 Sol, Sonnet 5.5, Gemini 4 Pro and Opus 5.5; the author finds Sonnet 5.5 strongest on detail and Gemini 4 Pro's moonlight lighting impressive.

The post rounds up three leaks: a Gemini 4 Pro checkpoint apparently on LMArena under the name Gemini 3.8 Flash, Claude Haiku 5.5 in partner testing, and GPT-6 Cyber preparing for release.

A tester says Gemini 4 Pro looks weak next to Opus 5.5 and GPT-6 Astra, with some abilities closer to strong open models like GLM and Kimi than to the frontier leaders.

A Chinese-language post claims Gemini 4 Pro will be released tomorrow, framing it as a gift for the country, in an unconfirmed launch rumor.