Muse Spark 1.1: Drop-In Agent, Self-Managed Context

On July 9, 2026, Meta's Superintelligence Labs shipped Muse Spark 1.1 along with the first public Meta Model API. The headline that got developers talking was not another benchmark chart. It was one sentence in the docs: the API speaks both the OpenAI Chat Completions format and the Anthropic Messages format out of the box. Your existing agent moves to Muse Spark by changing a base URL and a key.

The second thing worth your attention is quieter but bigger for real agents: Muse Spark 1.1 ships a 1,000,000-token context window that the model manages itself. Instead of you writing summarization glue to stay under a limit, the model decides what to keep active, what to retrieve from earlier in the session, and what to compress once a conversation runs long. For multi-step tool-use agents that accumulate thousands of tokens of tool output, that changes how you write the loop.