Meta released Muse Spark 1.1 — multimodal agent with 1 million token context

Meta Superintelligence Labs on July 9, 2026 unveiled Muse Spark 1.1 — a multimodal model designed for complex agentic tasks with a context window of 1 million tokens.

What Muse Spark 1.1 Can Do

The model was engineered for multi-step workflows where an agent orchestrates multiple applications and external services simultaneously. Muse Spark 1.1 can function as either the primary orchestrator of a multi-agent system or a specialized sub-agent within one.

• Context window — 1 million tokens
• Launch — July 9, 2026, public preview version via Meta Model API
• Available in "Thinking" mode in the Meta AI application and on meta.ai
• Supports images, video, and audio
• API compatible with OpenAI format

Multimodal capabilities include code generation from visual input, creating detailed descriptions for images and videos, and executing complex workflows with simultaneous use of multiple tools.

How Does the Model Handle Code and Computer Control?

On real-world tasks with large corporate codebases, Muse Spark 1.1 demonstrated significant improvement over its predecessor: diagnosing bugs, implementing features in existing systems, and performing large-scale code migrations. Particularly strong results were recorded on web application development tasks and multi-turn question-answer scenarios.

In computer control scenarios, the model is not limited to step-by-step click automation — it determines on its own when to write a script versus when to interact with the interface directly, and generates action packages in parallel. Amjad Masad, CEO of Replit, commented on the launch this way:

"What's most impressive about

Muse Spark is how much is packed into a single model: 1 million token context, full multimodal input (images, video, PDF), built-in search with citations, strong reasoning, top-tier code capabilities — especially in frontend and design, structured output, and parallel tool calls."

Where to Run the Model

Developers gain access to Muse Spark 1.1 through the public preview of Meta Model API with an OpenAI-compatible interface — migration from other providers is simplified. End users already have access to the model in "Thinking" mode in the Meta AI application and on meta.ai.

Before release, Meta conducted a safety assessment using the Advanced AI Scaling Framework, covering risks in chemical and biological weapons, cybersecurity, and loss of AI control. Based on the results, the model was deemed to operate within safe parameters and resistant to jailbreaks and prompt injection attacks.

What This Means

Meta is entering the market for advanced agentic AI with a competitive offering: million-token context, open API, and full multimodality — a direct challenge to OpenAI and Anthropic in the enterprise agent segment.

Frequently Asked Questions

How is Muse Spark 1.1 Different from Other Agentic Models?

According to Meta, the model combines a 1 million token context window, native multimodality (images, video, PDF, audio), and parallel tool invocation in a single system. CEO Saud Rizwan particularly emphasizes the attractive balance of capabilities and cost for real-world agentic workloads.

Do I Need a Separate API Key from Meta?

Yes — Muse Spark 1.1 is available through Meta Model API, which is an independent provider not affiliated with OpenAI or Anthropic. The interface is compatible with OpenAI format, which simplifies migrating existing code to a new provider.

*Meta is recognized as an extremist organization and is banned in Russia.