Meta CEO Mark Zuckerberg has announced on X that Muse Ember, Muse Spark 1.3 and Muse Wonder are now available on Google Cloud's Model Garden managed APIs. The announcement brings Meta's AI models to Google's cloud platform for the first time.
Google CEO Sundar Pichai replied to the announcement, saying: "Congrats and welcome to the model garden, excited to serve our Google Cloud customers!" The exchange between the two chief executives underlines how closely the rival companies are now working together on AI infrastructure.
Muse Ember, the smallest model
Muse Ember launched on 9 November 2026 as Meta's smallest and most cost-efficient model. It is built for repetitive, high-volume agent work such as tool calls, coding subtasks, routing, extraction and long document parsing.
Ember is priced at $0.10 per million input tokens and $0.50 per million output tokens. It is multimodal with a 1M token context window, and Meta says it delivers intelligence comparable to models like Claude Haiku at roughly one-third the per-task cost.
Muse Spark 1.3 is the workhorse agentic model for long-horizon, multi-step workflows. Muse Wonder is Meta's most capable frontier model, built for complex software engineering.
Where the models are available
The models are also available through the Meta Model API and Muse Code, and on Oracle Cloud. Google Cloud lists all three as Preview partner models on a global endpoint.
Meta Superintelligence Labs developed Ember, and Meta says Ember was trained with Muse Wonder Full. Zuckerberg summed up the lineup with the line: "Muse Wonder plans, Muse Ember handles the volume."
For developers, the arrival of the three models on Google Cloud means they can now run Meta's AI models alongside Google's own offerings on the same platform, with managed APIs handling the infrastructure.
