Meta has released Muse Spark 1.1, an upgraded multimodal AI model designed for agentic tasks. The company made the announcement on Thursday and opened a public preview to developers.
The new model focuses on computer use, coding and multimodal understanding. It features a 1-million-token context window and can orchestrate multiagent systems to handle complex projects with reduced latency.
Muse Spark 1.1 is available in Thinking mode inside the Meta AI app and on meta.ai. Developers can access it through the new Meta Model API.
Meta said the model manages long sessions by remembering prior actions and compacting context as needed. It also supports image and audio input for tasks such as interpreting screen content.
The release came from Meta's Superintelligence Labs, led by former Scale AI chief executive Alexandr Wang. The company highlighted extensive safety testing that showed resistance to jailbreaks and prompt injections.
