Meta’s Muse Spark 1.1 upgrade aims to sharpen agentic reasoning across coding and tool use

Meta Superintelligence Labs has released Muse Spark 1.1, an upgraded multimodal reasoning model designed to handle complex agentic tasks. The new version focuses on improvements in coding, tool and computer use, and multimodal understanding, according to a post on ai.meta.com. The release is part of Meta’s broader push toward what it calls “personal superintelligence,” as noted by Hacker News.

ai.meta.comMeta Superintelligence Labs launches Muse Spark 1.1 with enhanced multimodal reasoning for agentic tasks

“Muse Spark 1.1 is a multimodal reasoning model built for agentic tasks, with major gains in coding, tool and computer use, and multimodal understanding.”

The company positions the model alongside Muse Image, suggesting a suite of capabilities aimed at enabling AI systems to act more autonomously on behalf of users. Hacker News reported that the announcement is part of Meta’s “broader vision for personal superintelligence,” though specific use cases were not detailed.

Meta’s emphasis on agentic tasks reflects a growing industry trend toward models that can not only generate content but also execute multi-step actions, such as controlling software interfaces or writing and debugging code. The improvements in tool use imply the model is better equipped to interact with external APIs and digital environments, a key requirement for practical automation.

While benchmarks and specific performance figures were not disclosed in either announcement, the upgrade signals Meta’s intent to compete in the rapidly evolving multimodal reasoning space. Hacker News noted that Muse Spark 1.1 is positioned alongside Muse Image as part of a larger ecosystem, though the company has not yet outlined a public roadmap for broader availability.