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

A user had Gemini 4 Argon generate a 3D guide for peeling unripe plantains and an interactive 3D interior design site with color swaps, quotes, and WhatsApp handoff.

The author expected Gemini 4 Argon's momentum to overtake Claude Opus 5.5, but concludes Opus 5.5 is holding the top spot without much trouble.

A user grabbed a three-month Google AI Pro discount at about $1.42 per month mainly to try Gemini 4 Argon and asks existing users whether it is worth it.

An Ultra subscriber burned through the Gemini 4 Argon weekly quota in under two days in Antigravity despite using it far less than GPT-6.1 Sol.

A post claims Gemini 4 Argon converts dense PDFs into animated, explorable apps rather than static summaries.

The poster says to skip other models for creative writing and just use Gemini Argon for its human touch.

The post asks how Anthropic pulled ahead, noting Google is reportedly pushing RSI loops with Gemini 4 Argon while OpenAI races toward Astra 6.1.

The author chatted with an AI scam bot posing as Musk, got it to make images, then asked it to test Gemini 4 Argon and was lectured instead.

One test round puts Gemini AI Pro quota at roughly 1/7 of ChatGPT Plus (GPT-6 Astra) and 1/17 of Claude Pro (Opus 5.5), prompting a call for Google to raise limits.

The author runs the same prompt and shows Gemini 4 Argon's result, crediting Antigravity for the output.

A discussion post rounds up recent model releases including Gemini 4 Argon, GPT-6 Sol and Claude Opus 5.5, asking which builders use most.

A short post says people who doubted Google were proven wrong by Gemini 4 Argon, without offering any tests or data.