Meta has updated its Muse Spark multimodal reasoning model to version 1.1 and is making it available to developers through a public preview API, Axios reported. Meta’s own materials describe Muse Spark as a natively multimodal model aimed at reasoning, tool use and multi-step tasks.
Meta has updated Muse Spark to version 1.1 and opened developer access to the model through a public preview API, according to Axios.
Axios reported that Muse Spark 1.1 is focused on coding and longer agentic tasks, and that developers can access the model through a new Meta Model API with stated token pricing. The update expands access beyond Meta’s earlier private preview approach.
Meta’s own Muse Spark materials describe the model as a “natively multimodal reasoning model” from Meta Superintelligence Labs. In its original announcement, Meta said Muse Spark was designed to work across modalities and support capabilities such as tool use, visual chain of thought and multi-agent orchestration.
Meta’s Muse Spark product page also presents the model as part of the company’s broader push into AI assistants and reasoning systems. Axios reported that the 1.1 release is available in “thinking mode” in the Meta AI app and on Meta’s Muse Spark web experience.
The release puts Meta’s proprietary model work in more direct competition with other AI providers that offer developer-facing model APIs. By making Muse Spark 1.1 available through an API, Meta is giving developers a route to test the model in their own products rather than only through Meta-operated consumer surfaces.
Axios said the updated model is aimed at coding and longer-running agentic work. That positioning is notable because many AI companies are trying to move assistants from short prompt-and-response interactions toward workflows that require planning, tool calls and multi-step execution.
Meta’s earlier announcement framed Muse Spark around similar goals. The company said the model supports tool use and multi-agent orchestration, features typically associated with systems that can break tasks into parts, call external tools and coordinate multiple reasoning steps.
According to Axios, Meta is launching a public preview of the Meta Model API for developers. Axios also reported that the API includes token pricing, indicating that Meta is treating Muse Spark access as a developer platform rather than only a research demonstration or app feature.
Meta’s own announcement described earlier access as a private API preview. The move to a public preview, as reported by Axios, suggests a broader test of demand from developers and businesses.
The company has not, in the provided source material, disclosed enough detail to independently evaluate Muse Spark 1.1’s performance against competing models. Claims about improved coding ability or longer agentic tasks should therefore be read as Meta’s positioning and Axios’s report of the update, rather than as independently verified benchmark conclusions.
Meta has taken a mixed approach to AI model distribution, with some models released openly and others offered through company-controlled products and services. Muse Spark appears to sit in the latter category, with access through Meta AI, Meta’s web experience and the new developer API.
For developers, the practical significance will depend on pricing, reliability, latency, usage limits and how well Muse Spark 1.1 performs in real applications. For Meta, the release is another step toward turning its advanced AI models into products developers can build on.
Meta has updated Muse Spark to version 1.1 and opened developer access to the model through a public preview API, according to Axios.
What changed Axios reported that Muse Spark 1.1 is focused on coding and longer agentic tasks, and that developers can access the model through a new Meta Model API with stated token pricing.
The update expands access beyond Meta’s earlier private preview approach.
Continue reading
