Meta's Muse AI models now available on Google Cloud
What's the story
Meta's advanced AI models, Muse Ember, Muse Spark 1.3, and Muse Wonder, are now available on Google Cloud. Meta CEO Mark Zuckerberg announced the move on X, drawing a response from Alphabet CEO Sundar Pichai, who wrote, "Congrats and welcome to the model garden, excited to serve our Google Cloud customers!" The move marks a significant step in cross-company collaboration as Google Cloud will host and serve Meta's proprietary AI to enterprise customers along with its own Gemini offerings.
Model details
Muse Wonder
Muse Wonder is touted as Meta's most capable frontier model, designed for complex software engineering. It offers an optimal balance of speed, reliability, and deep persistence.
The model enables sophisticated multi-step workflows across browsers, desktop and mobile apps with native computer use and long-context understanding.
It also provides state-of-the-art coding, orchestration, multimodal reasoning capabilities, along with sub-agent delegation, parallel tool calling and agentic memory features.
Model features
Muse Spark 1.3
Muse Spark 1.3 is a workhorse agentic model available on Google Cloud for long-horizon workflows. It offers an optimal balance of intelligence, speed, and scalable performance.
The model enables the creation of sophisticated AI agents that complete multi-step tasks across professional and personal workflows with native tool calling and MCP server integration capabilities.
It also provides coding, reasoning, and autonomous execution capabilities with in-series specialization optimized for high-volume production environments on Gemini platform.
Model capabilities
Muse Ember
Muse Ember is an efficient, multimodal execution model available on Google Cloud for repeated everyday work. It offers an optimal balance of capability, throughput, and cost-efficiency.
The model provides comprehensive text, image, audio and video processing capabilities to help developers match task complexity with model capability across enterprise pipelines.
It also powers high-volume task routing and serves as the execution layer for complex multi-agent orchestration while analyzing multimodal media.