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

A post demonstrates using REA with Gemini 4 Argon to decompile and rebuild a target product through /recon, /analyze, /blueprint and /rebuild, noting most people still lack access to Argon.

A Spanish post compares Meta's Muse Wonder and Wonder Full with Gemini 4 Argon: DeepSWE 1.1 coding scores are nearly tied (77.9% vs 78.8%/79%), and Argon costs $2/$10 per million tokens until prices double after late November.

A user reports that three prompts with Gemini 4 Argon consumed 16% of their weekly usage limit and asks whether others are seeing the same issue.

An official example shows uploading a photo with rough measurements and using Canvas in Gemini to build an interactive 3D CAD studio with live sliders and STL export for 3D-printed furniture accessories.

An evaluation reports Gemini 4 Argon reaches 81.9% on RoboVista robotics understanding, beating GPT-6 Astra and Claude Fable 5.1 at the lowest eval cost.

A user combined Argon, Nano Banana and Veo in the Gemini app to turn a dense academic paper PDF into a full-color animated story, pitched as a new way to communicate research.

A Persian-language post says Google used Gemini 4 Argon with Nano Banana and Veo to turn a complex academic PDF into a full-color illustrated and animated storybook.

In a demo video, Gemini 4 Argon handles the kite physics simulation, Nano Banana 2 lite generates the visuals, and Lyria 3 provides the music.

Mike Knoop says Argon is only the second model family to effectively max out ARC-AGI-3 using a provider adapter, with the same "backwards bend" showing up in action efficiency.

Given the same 15-second showreel brief, Gemini 4 Argon went sculptural with depth and particle geometry while Claude Opus 5.5 favored flat typography and readable beats, showing very different ideas of what a showreel should do.

A user had Gemini 4 Argon generate an airport simulator game video, using 15.5M tokens at $1.96 in output cost over 2 hours.

A user tagged Google and GeminiApp to complain that the Argon usage limit is bad and needs improvement.