Summary
Google has released Gemini 4 Neon, its third major model launch in about a week, and Paul J Lipsky shares his hands-on results from early access. Google positions Neon as a fast reasoning model that sits between Gemini 4 Hydrogen, built for high-volume everyday tasks, and Gemini 4 Carbon, aimed at the hardest long-horizon problems. It promises quality close to Carbon for coding, reasoning and production agents, at higher speed and lower cost.
The video notes that Neon is rolling out in the Gemini app, including on a $20-a-month Pro account, as well as in Google AI Studio and Antigravity. No official benchmarks were available at recording time, so Lipsky relies on practical comparisons.
He runs several blind side-by-side tests between Neon and Carbon with identical prompts at high reasoning settings. These include an image comparison, a 3D forest-foraging game, a top-down paintball capture-the-flag game built with Godot and Blender, and a Star Fox-style starfighter game. In several cases he preferred Neon’s visuals, while Carbon came out slightly ahead on polish in the starfighter game. The video closes with guidance on which Google model suits which kind of task.
📺 Source: Paul J Lipsky · Published November 18, 2026
🏷️ Format: Review







