Muse Spark 1.1 by Meta AI

About Muse Spark 1.1 by Meta AI

Muse Spark 1.1 is a multimodal reasoning model from Meta Superintelligence Labs, released as an upgrade to the earlier Muse Spark. It is built for agentic tasks, with stated improvements in tool use, computer use, coding, and multimodal understanding. The model is available now in Thinking mode on Meta AI and via a public preview of the Meta Model API.

Review

The 1.1 update focuses on making the model more capable as an agent-handling multi-step tasks that involve interacting with tools and interpreting visual input. Early user feedback points to sharper tool calling and better coding performance. A 1M-token context window and multi-agent orchestration are also part of the release.

Key Features

• Multimodal reasoning that combines text, images, and tool interactions for agentic workflows.
• 1M-token context window, which lets the model process very long documents or extended conversations.
• Multi-agent orchestration to speed up execution of complex, multi-step tasks.
• Thinking mode, the current access point on Meta AI and meta.ai where the model runs.
• Public preview of the Meta Model API, giving developers a way to integrate the model into their own applications.

Pricing and Value

Pricing details for the API have not been disclosed. During the public preview, developers can access the model, but eventual costs are not yet defined. The model is available without direct cost in Thinking mode on Meta AI's consumer interfaces, though usage limits may apply.

Pros

• Handles multi-step agentic tasks with less hand-holding, particularly in tool use and browser automation.
• 1M-token context window accommodates large codebases, long research papers, or extended dialogues.
• Multimodal understanding catches subtle details in charts and images, as noted by early testers.
• Coding and tool calling feel noticeably tighter than the previous version, with fewer incorrect API selections.
• Public API preview lets development teams test agentic integrations before committing to a paid tier.

Cons

• The API is in a public preview; endpoints and behavior may shift before general availability.
• It is not well suited for straightforward Q&A or simple chat interactions where a smaller, faster model would be adequate.
• Multimodal and agentic capabilities can increase response latency for lightweight queries compared to text-only models.

Developers building AI agents or applications that need to orchestrate tools, interpret visual data, and handle long contexts will find the 1.1 update worth testing. The public API preview provides a chance to evaluate its agentic performance before pricing is set. For general-purpose chat, however, the model's specialized capabilities may be more than what's needed.

Open 'Muse Spark 1.1 by Meta AI' Website