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

A user complains that OpenAI's day six brought a schedule reset instead of a new model, and predicts Gemini 4 Argon could catch up to or beat GPT-6 Astra.

Users say the existing Gemini 3.8 Flash endpoint may secretly be running Gemini 4 Argon, citing behavior changes as a sign of a quiet test.

A post asks whether people are waiting for the unreleased Gemini 4 Argon or sticking with Claude Opus 5.5 for now.

The poster says Gemini 4 Argon still isn't available to regular users and asks how much longer they are supposed to wait on day 11.

A post calls Gemini 4 Argon Google's new frontier model for coding, enterprise work and cyber defense, and asks whether one model can handle an entire workflow.

A leak claims requests to the existing Gemini 3.8 Flash are being routed to Gemini 4 Argon, suggesting Google may be quietly testing Argon through an older model before launch.

Recalling that Google tested Gemini 3.5 Pro with partners but skipped straight to Gemini 4, the poster notes Argon was announced on September 30 as coming soon and hopes it ships rather than becoming another 3.5.

The poster wants Gemini 4 released next week, arguing its benchmark impact fades the longer Google waits, and notes Fable 5.5 and OpenAI's Astra 6.1 are also due soon.

A user says they suspect Satya's security premise of assuming Opus 5.5, Sol 6.1 and Gemini 4 Argon all get compromised will be borne out over time.

A user is skeptical of the Gemini 4 Argon hype, saying it may deliver less than 10% of what people expect.

The author says Gemini 3.5 Pro never arrived in June and Gemini 4 Argon, announced September 30, is still limited to select users, asking what benchmark wins mean without access.

A video post says another codename, Gemini 4 "Carbon", is drawing attention away from Gemini 4 "Argon", hinting at more unrevealed members of the Gemini 4 family.