Vibe Check: Gemini 4 Carbon's Coding Outpaces Its Judgment

Gemini 4 Carbon is a capable model that picked a bad day to launch. The new model from Google DeepMind arrives today, one day after Anthropic’s Fable 5.2 once again raised our expectations for what we can hand off to AI.

For Gemini loyalists and people whose jobs keep them inside Google’s tools, Carbon is a win. It’s a much stronger coder than Gemini 4 Argon, and it makes better three-dimensional objects. It can also take on ambitious builds and keep working for hours with little supervision.

But there’s a Fable in the room. It’s shown us that an agent can work for hours and exercise good judgment along the way. That means its competitors have to do more than handle long assignments to impress us: They need to choose the right work, respect the limits of the assignment, and recognize when it’s finished. Carbon makes DeepMind more competitive on those jobs, though our testers disagree on how much supervision it needs. It’s a better model, but Fable has raised expectations for what a flagship should deliver.

Subscribers only

Only available for paid subscribers

Subscribe to read the rest of this article.

Start free trial →

Already have an account? Log in.

Disclosure: Google provided Every with early access to Gemini 4 Carbon. Google had no input on this review.

A virtual summit for the new, new CFO

The era of the builder CFO is here. They are vibe coding deal desks, wrangling MCPs to deliver never-before-seen automations, setting up agent swarms to triage invoices, and imagining other inventive ways to methodically deploy agentic intelligence across finance workflows.

Traceable by Chargebee is a virtual event dedicated to documenting this shift. Join CFOs, controllers, finance engineers, and founders as they share field notes on the rise of the AI-native finance org. Featuring leaders from Public.com, Campfire, Sphere, Every, and other frontier teams.

Save your spot

Want to sponsor Every? Click here.

Katie Parrott is a staff writer at Every. To read more essays like this, subscribe to Every, and follow us on X at @every and on LinkedIn.

Stop explaining AI to your team and start showing them. Every Agent does the work in Slack, where everyone can see it, so one good workflow quickly becomes the whole team’s.