Meta launches Muse Spark 1.1 API with 1M-token context for agents

Meta has released Muse Spark 1.1, a multimodal reasoning model from its Meta Superintelligence Labs, to developers through a public preview of the Meta Model API. Announced on July 9, the upgraded model is purpose-built for agentic AI workloads that require extended planning, task orchestration, and interaction with external applications and services. Muse Spark 1.1 supports a 1 million token context window, enabling it to process large codebases, lengthy documents, and complex multi-step workflows without losing coherence. The model handles multimodal inputs—text, images, and tools—and can be used for coding, computer-use tasks, and other multimodal agent applications.

The release marks Meta’s push to compete in the agentic AI space, where models must not only generate responses but also plan, reason, and execute sequences of actions across diverse systems. Developers can integrate Muse Spark 1.1 via the Meta Model API to build assistants that autonomously navigate interfaces, write and debug code, or manage workflows spanning multiple SaaS tools. With the large context window, Muse Spark 1.1 maintains state across long interactions, making it suitable for advanced orchestration use cases. Early access suggests Meta is targeting enterprise developers building autonomous agents for automation, research, and software development.