MODELS

Muse Spark

by Meta

metamuse-spark

Overview

Muse Spark is a Meta Superintelligence Labs model family described by Meta as focused on multimodal reasoning, tool use, coding, and agentic tasks.

Details

Muse Spark is described in official Meta AI sources as the first Muse-family model developed by Meta Superintelligence Labs. Meta’s launch material describes native multimodal reasoning, tool use, visual chain of thought, and multi-agent orchestration. Meta also describes Muse Spark 1.1 as a multimodal reasoning model available through the Meta Model API public preview and in Meta AI Thinking mode, with an evaluation report covering capability and safety evaluations under Meta’s Advanced AI Scaling Framework.

When to Use

Evaluate a Meta-hosted multimodal reasoning model for workflows involving text visual reasoning tool use or function calling. Compare Meta Model API public-preview options for developer experiments coding workflows or agentic applications. Review Meta’s published evaluation and safety materials when assessing advanced model risk and deployment readiness.

Getting Started

Read Meta’s Muse Spark launch page: https://ai.meta.com/blog/introducing-muse-spark-msl/
Review the Muse Spark 1.1 and Meta Model API announcement: https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/
Consult the Muse Spark 1.1 Evaluation Report before building against the model: https://ai.meta.com/static-resource/muse-spark-1-1-evaluation-report
Check current Meta Model API access requirements and availability because the supplied source describes API access as public preview.

Key Features

•First Muse-family model from Meta Superintelligence Labs
•according to Meta’s launch page.
•Muse Spark 1.1 is described by Meta as a multimodal reasoning model.
•Meta cites native multimodal reasoning
•tool use
•visual chain of thought
•and multi-agent orchestration for Muse Spark.
•The evaluation report states Muse Spark 1.1 is available to external developers through an API with agentic affordances including tool and function calling.
•Meta publishes safety and preparedness materials covering chemical and biological
•cybersecurity
•loss-of-control
•adversarial robustness
•and model-behavior risks.

Capabilities

•multimodal reasoning
•visual reasoning
•tool use
•function calling
•coding
•agentic workflows
•multi-agent orchestration

Last updated Jul 10, 2026