
What happened
Anthropic's Claude Fable 5.2 made a photorealistic parking garage scene using roughly 700,000 simulated brushstrokes, per Anthropic employee Jake Eaton. Each piece is a Python program of 1,000 to 2,000 lines driving a simulated paint engine.
Why it matters
It shows a language model handling long-horizon visual and spatial tasks through code, with no external image model or traced source images, an approach that may carry over to physical robots.
WHO IT HITSDesign and creative teams evaluating AI image tools may find a model that produces paintings via code rather than diffusion. Robotics researchers could also benefit, as Anthropic reports Fable 5.2 scored 64.9 percent on its RoboDojo Sim benchmark versus 29.8 percent for OpenAI's GPT-6 Astra.
Summaries like this, every morning.
Anthropic positions Claude Fable 5.2 not as an image generator but as a language model that plans and executes long sequences of code under physical constraints. In one experiment, Anthropic engineer Michael Graczyk asked it to paint a bird at sunrise using only code to simulate bristle physics and wet paint; the resulting 21-second video tracks the painting from exactly 111 strokes on a blank canvas. The model chose color code #30346a, a width of 90 pixels, and five slightly wobbling waypoints for the first stroke, saying it added the roughly seven-pixel wobble on purpose because a human arm doesn't draw a perfectly straight line. Between passes, it looked at the canvas and corrected its own mistakes, such as laying a bolder stroke over a too-thin bird's back.
Earlier attempts with the sibling model Opus 5.5 drew criticism for merely imitating existing art. Eaton responded by having Fable 5.2 develop its own artist personas with fictional backstories and consistent styles, and by redirecting brushwork with comments. The same step-by-step, self-correcting behavior appears outside painting: Anthropic says Fable 5.2 solved two open problems in black hole physics, designed three working electronic devices with circuit board manufacturer Diode, and used more than 1,000 sub-agents to program its own laptop operating system in eight days.
AI researcher Aris Zhu tested the model with real robotic arms against OpenAI's GPT-6 Astra, each painting Van Gogh's "The Starry Night." Anthropic reports a score of 64.9 percent for Fable 5.2 on its internal RoboDojo Sim benchmark, compared with 29.8 percent for GPT-6 Astra. Zhu's real-world comparison still shows limits in fine motor control: Astra painted faster, and so far only OpenAI's model has managed to stand a domino on end, by slowly loosening its gripper to use gravity as a pivot point.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
The one-day session on 12月9日 teaches the basics of building AI agents with Agent Builder and Copilot Studio th…

kickflow raised about ¥700 million through a third-party allotment of new shares, with SMBC Venture Capital an…

Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model evaluated across more than 200…

Folkbench announced a method that pairs current-events probes, which resist static caching, with model-fingerp…

A Zenn writer tested opencodex 2.100.2, a local gateway that translates between Anthropic Messages, OpenAI Cha…

In a trial of replacing 34 files across subagents, agents given the delegating-work Skill met 18 of 24 criteri…
