Gemini 4 Carbon vs Claude Fable 5.2: Coding Skill and Real Cost Compared

At a glance

This video compares Google's Gemini 4 Carbon and Claude Fable 5.2, released a day apart. It looks at results from city robot building to real code-fixing tasks.

What the video covers

1. 1

   It explains, using inference settings and public API prices, why two models with similar top scores in the Cursor evaluation had different costs.

2. 2

   It explains how Carbon's launch discount, cache use, and the harness (the framework an AI works within) affect what users are charged.

3. 3

   It checks whether the cheaper model always wins, using harder tasks such as a file system that must survive power outages.

4. 4

   It also covers how multiple agents (AI workers) split up work, along with math proof checking and circuit design cases.

Why watch it?

This is useful for developers and anyone considering adoption who wants to weigh both AI coding tool performance and cost.