Meta launches Muse Spark 1.1 and public Meta Model API

On July 9, 2026, Meta Superintelligence Labs announced Muse Spark 1.1 — a multimodal reasoning model built for agentic tasks, with major gains in tool use, computer use, coding and multimodal understanding. Alongside it, Meta opened a public preview of the new Meta Model API for developers; Muse Spark 1.1 is also available in "Thinking" mode in the Meta AI app and on meta.ai.

Key points

• A significant upgrade over Muse Spark, tuned for planning and orchestration across many apps and services; it zero-shot generalizes to new native tools, MCP servers and custom skills.
• Multi-agent orchestration: it acts as the main agent to plan and delegate execution across parallel subagents to reduce end-to-end latency.
• A 1-million-token context window that it actively manages — remembering actions, retrieving earlier work and compacting while keeping the steps that matter.
• Computer use: it decides when to write scripts to automate, when to interact directly, and generates batches of actions per step.
• Coding: stronger on large, complex codebases (diagnosing/fixing bugs, adding features, running migrations) and compatible with popular agentic harnesses (planning mode, subagent delegation, context compaction).
• Safety: evaluated under Meta's Advanced AI Scaling Framework; across Chemical & Biological, Cybersecurity and Loss-of-Control risk categories it operates within safe margins, with better resistance to jailbreaks and prompt injection.

FAQ

Where can I use Muse Spark 1.1? Via the Meta Model API (public preview) for developers, and in "Thinking" mode in the Meta AI app and on meta.ai. What's the main difference from the prior model? Large gains in agentic capability, computer use, coding on big codebases, and 1M-token context management.

See Meta's official announcement for details and the evaluation report.