Gemini 4 Carbon Review: Stronger Coder Than Argon, but Fable 5.2 Raised the Bar

every · x · 2026-11-14

Google DeepMind released Gemini 4 Carbon a day after Anthropic's Fable 5.2 reset expectations for flagship models.
Every's testing: Carbon codes much better than Gemini 4 Argon, makes better 3D objects, and can run ambitious multi-hour builds with little supervision.
Real case from Everycut: several models missed recurring timeline issues (cuts clipping speech); Carbon spotted them and autonomously opened a fix PR that was merged after review.
The catch: Fable 5.2 showed agents can work for hours and exercise judgment—choosing the right work, respecting limits, knowing when they're done. Testers disagree on how much supervision Carbon still needs.
Verdict: a clear win for Google loyalists, but "a better model" hasn't matched the bar Fable set.

Note: full review is subscriber-only; Google provided early access with no editorial input.

More from coding & agent