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

Replying to Logan Kilpatrick, a user says nobody cares about 3.8 Flash TTS and the community only wants Gemini 4 Pro.

Replying to Logan, a user complains that refreshing the feed brought no Gemini 4 Pro launch news and asks whether it is ever coming out.

In a reply to Logan Kilpatrick, a user jokingly demands Google release Gemini 4 Pro or open-source Gemini 3.8 Flash.

A user replies to Logan's post asking about Gemini 4 Pro, suggesting the community sees it as the next frontier model to launch.

After September speculation fell through, new reports point to a mid-to-late October rollout, with Google possibly waiting for the next frontier-model wave.

A user who opened the circulating Gemini 4 Pro link says it is actually Gemini 3 Flash without thinking mode, pushing back on the leak.

A screenshot shows the "Gemini 4 Pro" label is only a custom name rather than real model routing, casting doubt on the leak.

Under Google AI Studio's official account, a user says they don't want new features and are only waiting for Gemini 4 Pro to launch.

A user questions why Gemini 4 Pro can generate almost anything yet struggles to fix Antigravity, asking whether AI is actually getting smarter or just better at looking impressive.

A rumor says a Gemini 4 Pro checkpoint is being tested in Arena under the "Gemini 3.8 Flash" label, may reach Mythos-level performance, and could face GPT-6 Astra with an October launch.

A post says reported checkpoints show Gemini 4 Pro making a huge leap in 3D, SVG and frontend coding.

Users link the suddenly stronger gemini-3.8-flash on Arena to an internal codename Argon/G4P, claiming it beats Astra on coding, SVG, 3D and agent scores, though Google has not announced anything.