Gemini 4 Carbon vs Claude Fable 5.2!

Julian Goldie (0:00)

Gemini 4 Carbon vs. Claude Fabel 5.2.

Two brand new frontier models launched one day apart. So I put them on the same track. Five laps, five real business builds. Same prompt, same start signal, and a real clock running on both.

One of them shipped a broken app and told me it passed. The other one lost three laps on design. And the final score came down to the very last lap. Let's go.

Here's what just happened.

On November 9th, Anthropic released Claude Fabel 5.2.

One day later, Google said Gemini 4 Carbon is now available. Both are built for long jobs. Both read up to a million tokens of context. Fabel runs in the Claude apps and Claude code for everyone. Carbon is rolling out in anti-gravity and the Gemini app for Google AI Ultra subscribers and to developers in AI Studio. On paper, it's close.

On the Artificial Analysis Intelligence Index, at the same high setting, Fabel scores 62 and Carbon scores 60

Anthropic's own table puts Fabel ahead on Frontier Sui, 69.9 to 61.2.

But that same table has Carbon ahead on Automation Bench, 53.9 to 51.4.

Benchmarks are one thing. I wanted to see what happens when you hand each one a real job. So here are the rules.

Fabel runs in Claude Code. Carbon runs in Anti-Gravity, on its normal setting, one agent, no swarm. Setting this up takes seconds. In Claude Code, you type slash model and pick Fabel 5.2.

In Anti-Gravity, you click the model name under the message box and pick Gemini 4 Carbon. I left both on medium, which is the default. To be up front, Claude Code had my own design skill installed, so it pulled that in on four laps. Carbon ran as it comes.

Every lap is something a member of the AI Profit Boardroom asked about this week, and every lap gets the same four checks at the pit wall. Speed, does it work first try? Design, and did it follow the brief?

Lap 1, a member who sells websites posted this week. They want to build their own client system instead of a bloated CRM. Leads, follow-ups, bookings, and reviews in one place. So both models got this brief. Build a client hub for a web design agency. A leads inbox, automatic follow-up reminders when a lead goes quiet for two days, a week view calendar, and a reviews tab. Lights out, Fable finished in 8 minutes 36 12 sample leads, a banner that tells you three leads have gone quiet, live counters across the top, and every tab works. Carbon took 11 minutes 36 It even started its own checker agent to test the app, and the checker said everything passed. Then I opened it in a fresh browser. The leads list was empty. Every counter said zero, and the console showed an error on load. That's the most important check in this whole video. An app that looks finished, but breaks the first time a client opens it, is worse than no app at all. Lap 1 goes to Fable. And if you run an agency, this is the build to steal. Give one of these to every website client, and the follow up reminders alone stop leads going cold. If you want the exact prompts from this video, plus the finished builds, they're inside the AI Profit Boardroom. You get all five LAP prompts word for word, the client hub, the audit report, the 3D landing page, the model router, and the storefront, plus the cheat sheet I'll show you at the end for which model to use for which job. You also get my faceless video skill, the one that made this video, and a community of people building the same things, so when something breaks, you're not stuck on your own. The link is in the comments and the description, or go to aiprofitboardroom.com.

Lap 2 A member who works as a consultant is running their first two AI audits on-site next week. They'll sit with the owner and the team and watch how work actually gets done. So I wrote up messy notes from an audit of a small coaching business. One founder, two part-time assistants, about 140 clients, onboarding done by hand, recaps written every Monday. The brief? Turn these notes into a polished audit report. A scorecard, the top eight automation opportunities on an impact versus effort chart, and a 30, 60, 90-day plan. Fable finished in 6 minutes 20 A clean green report with a big headline, a scorecard, and a matrix that ranks every opportunity. Carbon took 8 minutes 9 Its report is just as clean, and it went further on the facts. It flags assumptions all the way through, and adds a box explaining where each number came from. Both of these are reports you could hand to a client today.