Meta Superintelligence Labs Launches Muse Spark 1.1 and Public API

Meta Superintelligence Labs has launched Muse Spark 1.1, a multimodal reasoning model engineered for agentic tasks, alongside a public preview of the Meta Model API, marking a significant strategic pivot for the tech giant. Released on July 9, 2026, this move represents Meta's inaugural foray into offering its proprietary models via a paid developer API, departing from its traditional open-weights approach. The model distinguishes itself with a formidable 1,000,000-token context window, actively managed through compaction and delegation, enabling it to handle extensive, tool-heavy workflows.

Muse Spark 1.1 is positioned as Meta's most capable model for real-world coding and agentic applications, demonstrating substantial improvements over its predecessor in tool use, computer interaction, coding, and multimodal understanding. While benchmarks show it leading in professional and scaled tool-use categories like JobBench and MCP Atlas, it currently trails rivals such as Opus 4.8 and GPT-5.5 in pure coding and certain multimodal reasoning tasks. Crucially, Meta is adopting an aggressive pricing strategy, offering access at $1.25 per million input tokens and $4.25 per million output tokens, coupled with $20 in free credits, directly challenging established players like OpenAI and Anthropic. The API's compatibility with the OpenAI SDK further lowers the barrier for developers to integrate Muse Spark 1.1 into existing systems.

This launch signifies Meta's serious intent to monetize its advanced AI research and compete directly in the burgeoning market for AI agents. The focus on "agentic tasks"—models that can plan, orchestrate, and execute complex actions across various applications—points towards Meta's vision of "personal superintelligence". While the competitive pricing and OpenAI SDK compatibility could attract a broad developer base, the ultimate success of Muse Spark 1.1 will hinge on its real-world performance and the reliability of its expansive context window, an area where many large models face challenges. This strategic shift could redefine Meta's role in the enterprise AI landscape, potentially driving down costs and accelerating innovation in agentic AI development.