AboutAI ToolsCursorClaude CodeLovableBolt.newAI tool reviewsTutorialsVibe codingAI agentsMobile apps with AIAll tutorialsBlogSponsorWork With MeJoin 285K+ Subscribers

15 November 2026 · 2 min read

Claude Fable 5.2: 3D Game From a Screenshot and More – My Real-World Tests

I put Claude Fable 5.2 through six real 3D and design projects, from a racing game built from a single screenshot to Blender character rigging. Here’s what worked, what didn’t, and the full spec breakdown.

Claude Fable 5.2: 3D Game From a Screenshot and More – My Real-World Tests

What I Tested

I put Claude Fable 5.2, Anthropic’s newest model, through six real-world 3D and design projects to see if it lives up to the claim of being the best model yet for 3D worlds, games, and motion graphics.

The Six Projects

Highway racing game from a screenshot – I gave the model a single screenshot and it produced a playable highway racing game. That was one of the standout results.

3D Giza plateau – From a single prompt, it generated a 3D scene of the Giza plateau. This worked well and showed its ability to understand spatial layouts.

Low-poly Manhattan city in three.js – Another single-prompt test. The model built a low-poly version of Manhattan that ran in the browser using three.js.

Finance app UI – I asked for a user interface for a finance application. The output was a functional UI design, demonstrating its grasp of layout and component structure.

3D character in Blender – Using Blender MCP inside Claude Code, I had Fable 5.2 recreate, rig, and animate a realistic 3D character. This was a more complex task, and the results were mixed – some parts worked, others didn’t. I show both in the video.

Adjustable MacBook stand – I also used Blender MCP to design and simulate an adjustable stand for a MacBook. Again, the model showed promise but wasn’t perfect.

I walk through each project in the video, highlighting where Fable 5.2 excelled and where it fell short.

Specs and Pricing

After the demos, I break down the full specification sheet:

  • Context window: 1 million tokens
  • Maximum output: 128,000 tokens per reply
  • Thinking is always on, with five effort levels: low, medium, high, extra high, and max
  • Speed: up to 30% faster than Fable 5.1; a fast mode can make it up to 2.5x faster again
  • Agent coordination: can manage over 1,000 sub-agents on a single job
  • API pricing: $8 per million input tokens and $40 per million output tokens – 20% cheaper than Fable 5.1. Fast mode costs $16 / $80 per million tokens. For comparison, Opus 5.5 is $4 / $20.

I also discuss when you should choose Fable 5.2 over Opus 5.5.

Who Should Watch

If you’re a developer or founder building 3D experiences, games, or design tools, this video gives you a practical look at what Fable 5.2 can actually do – and where it still needs work. The spec breakdown will help you decide whether to adopt it for your own projects.

Watch the full video above to see all the results and my detailed analysis.

Sources

  1. Claude Fable 5.2 Built a 3D Game From One Screenshot - Claude Fable 5.2 is INSANE
  • claude fable 5.2
  • 3d game development
  • anthropic
  • three.js
  • blender mcp

Watch the full video on YouTube

Written with AI from the video above.

Learn to build with AI

Hands-on tutorials on Cursor, Claude Code, AI agents and more.