Meta said in its September 24 Connect developer recap that the Meta Model API is generally available worldwide, expanding a service previously offered in public preview. The announcement also moves Muse Code out of beta on Windows and broadens access through cloud partners and a local model option.
Advertisement
Meta says enterprise support and compliance are offered through its first-party API and partners. The recap does not provide a complete rate card or eligibility conditions for each offering.
Muse Spark 1.3 is available through Oracle Cloud AI Platform, according to Meta. Google Cloud access remains in private preview. Spark 1.3 itself was introduced on September 2; Connect’s news is broader access and developer tooling, not a new launch of the same model.
Coding, media and local deployment
Muse Code is designed for extended engineering tasks, with parallel workers writing and testing code in isolated workspaces. Its move out of beta and onto Windows concerns the coding-agent product, separate from the availability of a model endpoint.
The recap also points developers to Muse Image for image generation and Muse Voice Transcribe for speech-to-text work. These serve different input and output needs from a reasoning model. Meta links the product family and development resources from its developer portal rather than describing them as one interchangeable API feature.
For local deployment, Meta describes Muse Glimmer as an open-weight, 30-billion-parameter model designed for a single GPU or Mac Mini. The recap does not specify its license, checkpoint requirements or deployment configuration.
Meta also highlights the Muse Connector Platform, intended to connect developers’ services with tasks performed through Muse. This adds an integration path alongside choosing a hosted endpoint or running weights locally.
The developer recap provides a starting point at dev.meta.ai and links to documentation and cookbook recipes.