Meta's Superintelligence Labs ships its first image model, and it works like an AI agent.
Meta has released Muse Image, the first image generation model from its Superintelligence Labs. It's the first release of its kind since the company restructured its AI lab under Chief AI Officer Alexandr Wang, backed by billions in new investment.
Muse Image doesn't map prompts directly to images. Instead, it works as an agent, similar to OpenAI's GPT Image 2. The model calls tools to produce more accurate results and refines its own outputs.
Those tools include writing and running code to generate things like correct diagrams, scannable QR codes, animated GIFs, websites, or interactive games. A web search feature lets the model ground images in current facts and real-world references, which Meta says improves accuracy on knowledge-heavy prompts.
Meta's Muse works as an agent, much like OpenAI's latest model, iterating on results to improve them. | Image: Meta
