Meta’s Muse Spark 1.1 pushes agents into public preview

What happened

Meta released Muse Spark 1.1 and opened a public preview of its Model API. The model is designed to work across text and images while using tools to complete multi-step tasks. A version of its reasoning is also appearing in Meta AI under a feature called Thinking.

Why it matters

Agent-style systems are moving from demos into products people can actually test. That changes the conversation. The challenge is no longer only whether a model can plan a task, but whether it can keep track of context, recover from a mistake and ask for help at the right moment.

Meta’s reach gives the launch weight. If the API becomes easy to use and competitively priced, it could widen the field for developers who do not want to build around a single model provider.

What is confirmed

Meta describes Muse Spark 1.1 as multimodal and agentic, and says the Model API is in public preview. The company has published technical results, but independent evaluations and sustained production experience remain limited at launch.

What remains unclear

The details that make or break an agent — stable tool use, predictable cost, permissions and recovery from failure — are best learned in real deployments. Developers should treat the preview as a structured trial and keep consequential actions behind human approval.