Meta's three new Muse AI models arrive on Google Cloud's Model Garden in preview

Meta has made three of its new AI models, Muse Wonder, Muse Spark 1.3 and Muse Ember, available on Google Cloud's platform Model Garden. Meta CEO Mark Zuckerberg announced this on the social media platform X. He wrote, "Muse Ember, Muse Spark 1.3 and Muse Wonder are now available on Google Cloud."

Replying to Zuckerberg's post, Google CEO Sundar Pichai wrote, "Congratulations and welcome to our Model Garden! We are eager to serve this advanced technology to our Google Cloud customers."

According to Google Cloud, all three models are currently in preview. Their biggest feature is a context window of up to 1 million tokens. In simple terms, a context window means the model can take in information from hundreds of pages, thousands of lines of code or large documents at once, and remember their context without losing track.

Muse Wonder is regarded as the most advanced, high-end model in Meta's AI series. It is built to handle complex software engineering workflows, simultaneous updates to many files and long-running coding projects. It breaks a large, difficult task into smaller parts and hands them to different AI agents so that the work is finished sooner. It also has parallel tool calling, which allows multi-step tasks on browsers, desktops and mobile apps to be automated together. It is also said to be adept at examining thousands of lines of code or legal documents and drawing precise results.

Muse Spark 1.3 is made especially for tasks that run in many steps and over a long time. Developers will be able to use it to build their own AI agents for different applications. The model fully supports the Model Context Protocol (MCP) and native tool calling, so AI agents connect easily to different software tools and APIs. Its use is described as effective in building software, API integration, personal assistants and enterprise-level workflow automation.

Muse Ember is meant for companies and developers who want to finish routine, repetitive tasks at lower cost and in less time. According to Meta's claim, the model runs with very low latency and is available at a much lower cost than larger models. It supports both text and image or screenshot inputs. It is described as well suited to solving simple coding problems, summarising documents and running routine agent workflows.

Google Cloud's enterprise customers will now be able to use these new Meta models alongside Google's own Gemini models. This means companies will not have to depend on a single AI model. They can choose the right model from Gemini or Meta's three Muse models according to the needs of their work. For both small and large workloads, they will also be able to choose Wonder or Ember depending on the size of the task.