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

A developer ships a 600M-parameter Persian vision model built on Nemotron 3.5 Parse, claiming it competes with Gemini Argon on their benchmarks.

A developer lists what one Gemini API key enables for agents—image/video understanding, audio, generation, TTS, cheap transcription and classification—and says Argon 4 is creative and uncorrelated with Claude/OpenAI models, near frontier but still behind Opus 5.5 for daily use.

A user asks whether the €5.50/month student Google AI Pro plan with Gemini 4 Argon, paired with Codex, is worth it for AI agent coding.

FrontierCode 2.0 separates model tiers more sharply: Muse Wonder 1.0 and Gemini 4 Argon top DeepSWE v1.1 but land mid-pack on FrontierCode 2.0.

The poster argues that Gemini 4 Argon alone won't make Google's Antigravity mainstream among developers and asks for agreement.

After updating Antigravity, users can select Gemini 4 Argon High marked "New" in the model list; those who don't see it should update the IDE.

ZenMux's Halloween promo lets users who spend $1 in cash PAYG daily open a model blind box; the 16-model pool includes 5M Gemini 4 Argon tokens, with rewards up to $500.

The author argues the next Flash won't be called 4 Flash, that gemini-4-flash-preview was added in June and deleted in August, and gemini-3.8-flash remains the active version.

StatsWire checked Google's python-genai SDK git history and found the gemini-4-flash-preview line added June 10 and deleted August 10, absent from main for 2.5 months with no new info today.

A user reports Gemini 4 Argon blocked nearly all invalid GPT code reviews, catching 13 erroneous items.

A leak says Gemini 4 Flash is in internal testing as "Hydrogen", a gemini-3.8-flash-router model ID appears in Antigravity, and launch may come this week or next.

Googlers are said to be testing a Carbon checkpoint in the internal Jetski coding tool, with one calling it Opus 5.5-like, and it reportedly appeared on Chatbot Arena with 3D praise.