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

A user complains that Gemini 4 Argon usage drains too fast on the $200/month plan and expects things to get worse next year once the promo pricing expires in December.

Running an old prompt through Argon High, the car shape is more accurate but colors are washed out, textures float, and the garage scene is too basic, leaving the author unsure whether to switch from Opus.

Gemini 4 Argon is now available in the Antigravity desktop app and CLI for Google AI Ultra individual subscribers, plus enterprise API and Gemini Enterprise customers.

A hands-on report says Argon is far ahead of earlier Gemini models and has a big-model feel, but is still beaten by Sonnet 5.5 on the same voxel prompt.

The poster says they just got access to Gemini 4 Argon and that it looks notably strong in science and 3D intelligence based on demos.

An Arabic-language account says Gemini 4 Argon targets complex multi-step tasks with a million-token context and is available to AI Pro/Ultra subscribers and developers.

Cursor says Gemini 4 Argon is now available, scoring 52.8% on CursorBench at default effort, on par with Opus 5.5, priced at $2/M input and $10/M output tokens.

A 3D/AI creator tests Gemini 4 Argon with a Rick Roll benchmark, with a video asking whether AGI can already be felt.

A post asks whether anyone has already tried Gemini 4 Argon, inviting replies from people who have tested it.

A leak says Gemini 4 Argon is now available to Ultra subscribers with a 1,048,576-token shared context, $2/$10 per million input/output tokens, doubled rates past a million tokens, and 95% off cached input.

Gemini 4 Argon is the new state of the art on SimpleQA Verified, ending Gemini 3.1 Pro's seven-month run at the top.

With the same prompt and no reference image, Argon took 6 minutes and $1.16 while Opus 5.5 took 26 minutes and $4.89, with the video revealing the winner.