📣 Send us your press release
Site updates every 15 minutes
Technology

Anthropic introduces Claude Fable 5.2 with code-driven image generation

Anthropic’s latest model, Claude Fable 5.2, produces visuals by writing Python code to simulate brushstrokes rather than using standard image diffusion models.

16 November 2026

Anthropic has released Claude Fable 5.2, a model that departs from traditional generative AI image methods by employing a programmatic approach to visual creation. Instead of using diffusion models to transform noise into pixels, Fable 5.2 writes and executes Python code that simulates the physical act of painting on a virtual canvas.

The model generates a series of stroke instructions, effectively acting as an engine that manages brush pressure, paint wetness, and canvas texture. Throughout the process, the model monitors its own progress, allowing it to identify imperfections and apply corrective brushstrokes to refine the final output. Each work is the result of thousands of autonomous stroke commands generated within a 1,000 to 2,000-line script.

Beyond artistic applications, Anthropic suggests this iterative, self-correcting capability is intended for complex technical tasks, including robotics and scientific research. In testing environments, the model demonstrated the ability to coordinate multiple sub-agents to solve long-term problems. While Fable 5.2 has outperformed some competitors in simulated robotics benchmarks, the company acknowledges that fine-motor control for physical hardware remains an area for further development.

Original source: ithome.com