Robots Atlas>ROBOTS ATLAS
Muse (Meta)
MetaActive

Muse (Meta)

Muse is Meta's AI system built for developers, combining a model family and a toolset into a coherent platform for building AI agents and AI-assisted applications.

At its core is the Muse Spark model family (latest 1.3), trained for long-horizon, agentic workflows and coding tasks. It is complemented by Muse Code (a terminal-based, multi-agent coding tool integrated with Muse Spark), Muse Glimmer (a local agent model), Muse Image (visual reasoning and planning), and Muse Voice Transcribe (speech-to-text with speaker identification).

The system stands out for long-context reasoning, multimodal perception (video, images, documents), and agentic task execution — it tracks context and prior results, works through messy or conflicting inputs, and asks for input when needed. Access is via the Meta Model API (a self-serve platform) and partners such as OpenRouter, with compatibility for OpenAI SDK-compatible tools.

APICLISDKUsage-basedPaid
Released 5 Aug 2026Updated 22 Sept 2026
Powered by
Muse Spark
Muse Spark
Model profile
Muse Spark 1.1
Muse Spark 1.1
1 mln tokenow ctx
Model profile
Muse Image
Muse Image
Nie dotyczy (model image gen; wielo-obrazowe blendowanie zamiast okna tokenowego) ctx
Model profile

Overview

System type
Input
TextImageVideoDocumentsAudio
Output
TextCode

Access & Pricing

Access channels
API
Programmatic access via an API interface.
CLI
Access via command-line interface.
SDK
Developer library for embedding the AI system in other applications.
Subscription model
Usage-basedPaid

Developer access via the Meta Model API (self-serve, usage-based billing) and via partners like OpenRouter. Detailed rates depend on the model.

Capabilities

10 capabilities
Reasoning & Planning
Agency & Action
Communication & Multimodality
Task Specialization

Sources

1 source
Meta Muse — dev.meta.ai
Fetched: 22 Sept 2026