---
type: "post"
id: "meta-muse-wonder-muse-spark-1-3-and-790011"
slug: "meta-muse-wonder-muse-spark-1-3-and-790011"
title: "Meta Muse Wonder, Muse Spark 1.3 and Muse Ember now available in Google Cloud model garden"
description: "Google Cloud’s Model Garden page lists Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember as models developers can deploy through its platform."
retrieval_nugget: "Google Cloud’s Model Garden page lists Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember as models developers can deploy through its platform. The lineup offers different stated capabilities for engineering, long-horizon agent work and repeated multimodal tasks, giving Google Cloud teams another model set to evaluate within their existing deployment environment."
published_at: "2026-11-10"
updated_at: "2026-11-10"
record_date: "2026-11-10"
date_kind: "published_at"
topics: ["google"]
entities: ["Google Cloud","Meta","Muse Wonder","Muse Spark 1.3","Muse Ember","Muse Spark Open","Model Garden"]
related_patterns: ["agent-protocols-become-interoperability-layer"]
novelty_label: "new_fact"
what_changed: "Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember models are now available through Google Cloud Model Garden."
prior_refs: []
source_url: "https://cloud.google.com/products/model-garden/meta"
source_urls: ["https://cloud.google.com/products/model-garden/meta"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:meta-muse-wonder-muse-spark-1-3-and-790011"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/meta-muse-wonder-muse-spark-1-3-and-790011/","markdown":"https://newruntime.com/posts/meta-muse-wonder-muse-spark-1-3-and-790011.md","json":"https://newruntime.com/posts/meta-muse-wonder-muse-spark-1-3-and-790011.json"}
markdown_url: "https://newruntime.com/posts/meta-muse-wonder-muse-spark-1-3-and-790011.md"
---

# Meta Muse Wonder, Muse Spark 1.3 and Muse Ember now available in Google Cloud model garden

## Retrieval answer

Google Cloud’s Model Garden page lists Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember as models developers can deploy through its platform. The lineup offers different stated capabilities for engineering, long-horizon agent work and repeated multimodal tasks, giving Google Cloud teams another model set to evaluate within their existing deployment environment.

## Summary
Google Cloud’s Model Garden page lists Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember as models developers can deploy through its platform. The lineup offers different stated capabilities for engineering, long-horizon agent work and repeated multimodal tasks, giving Google Cloud teams another model set to evaluate within their existing deployment environment.

## What happened
Google Cloud’s Meta models page presents Muse Wonder, Muse Spark 1.3 and Muse Ember as the newest Meta models available through Model Garden. It describes Model Garden as a place to discover, customize and deploy both proprietary and open-weight models, with serverless APIs among the access options. The page links to individual model entries and documentation for Wonder, Spark 1.3 and Ember. It does not provide a launch date or a detailed account of regional availability.

The page positions Muse Wonder as Meta’s most capable frontier model for complex software engineering. Its listed capabilities include native computer use, long-context understanding, sub-agent delegation, parallel tool calling and agentic memory. Google Cloud identifies multi-file repository updates, software engineering and end-to-end workflow automation as example uses, and also mentions operating applications and reasoning over large codebases. Google Cloud says the model can work across browser, desktop and mobile applications. These are vendor descriptions rather than independently reported performance results.

Muse Spark 1.3 is described as a workhorse model for long-horizon workflows and high-volume production. The page says it supports native tool calling and MCP server integration, and lists coding, reasoning and autonomous execution among its capabilities. Example applications include multi-step workflow automation, tool-using applications and API or MCP pipelines. Google Cloud also describes it as suitable for personal and enterprise assistants and for dynamic, real-world tasks at scale. Google Cloud links to a dedicated model entry and documentation. It does not state a price, usage quota or measured latency for Spark 1.3.

Muse Ember is presented as a lightweight multimodal execution model for repeated everyday work. Google Cloud says it handles text, images, audio and video and supports coding, tool calling and document analysis. Its proposed uses include everyday coding subtasks, automated tool execution, document understanding and high-volume task routing. The page frames Ember as a complement to Wonder and Spark in a multi-agent orchestration setup, where it can keep longer-running workflows moving.

The same page lists Muse Spark Open separately as an open-weight model deployable on Google Cloud: an 816-billion-parameter mixture-of-experts model with 42 billion active parameters and a one-million-token context window. It accepts text, image, video and audio inputs across more than 100 languages, according to Google Cloud. The page also says Google Cloud users can access Llama 4 and Llama 3 model families through the portfolio, but this availability update specifically adds the three newer proprietary Muse entries to the listed choices.

## What the archive adds
New Runtime’s earlier coverage said Muse Spark Open could be deployed in an account a team controls through [Amazon SageMaker JumpStart](https://newruntime.com/posts/meta-s-muse-spark-open-now-available-in-831a6b/) on October 21, 2026. An earlier New Runtime note also examined its [Muse Model License](https://newruntime.com/posts/meta-s-muse-spark-open-page-on-hugging-9dd8ba/) on the same date. This is a new platform-availability fact for the related Muse family, not a restatement of either earlier note.

## Trend signal
This release weakly supports [agent protocols becoming the interoperability layer](https://newruntime.com/patterns/agent-protocols-become-interoperability-layer/): Google Cloud specifically advertises Spark 1.3’s MCP server integration as an agent-building capability. It also fits the archive’s observation that platforms can keep evaluation and management around swappable models separate from a single model choice. The listing establishes a vendor-described integration, not evidence of broader adoption or interoperability outcomes.

## Why now
Google Cloud groups models with different capability levels and workload emphases in one Model Garden portfolio. That makes the update relevant to teams that want to compare model choices without moving deployment, security and management work to a separate environment. Google Cloud does not explain why these models became available at this particular time, so the timing should not be read as a stated product-roadmap signal.

## Connections to other events
The new entries broaden the cloud deployment options around Muse Spark Open. New Runtime’s earlier coverage of its Amazon SageMaker JumpStart availability focused on putting an endpoint inside an account controlled by the team. Google Cloud now presents a related portfolio that includes proprietary Muse models, serverless APIs and managed deployment choices alongside Spark Open.

## Why it matters
For teams already building on Google Cloud, the page describes a route to evaluate Meta models without treating model selection and infrastructure as separate procurement decisions. The distinctions are useful as hypotheses for a routing strategy: reserve a more capable model for difficult engineering work, consider Spark 1.3 where tool use and longer workflows matter, and test Ember on repeated multimodal or document tasks. Those are sensible evaluation candidates, not demonstrated cost or quality advantages.

The page also makes an important distinction between the proprietary Muse models and Muse Spark Open, which it calls open-weight. Teams should not assume that open weights imply identical deployment terms, control or operational requirements across environments. A controlled trial should measure completion quality, failure recovery, latency and total serving cost on representative tasks. The Model Garden listing provides a starting point, but its capability descriptions alone cannot establish whether any model improves a production workflow.

## Limits and uncertainty
Google Cloud’s page supplies capability claims and suggested use cases, but it does not report independent evaluations, benchmark scores, measured reliability or comparative results against other models. It does not state token prices, endpoint costs, quotas, latency, regional availability or the operational requirements for deploying the open-weight model. The precise access conditions for each model may depend on the linked console entries. Teams should confirm those details directly before committing workloads; the listed capabilities should not be read as guarantees of task performance.

## Skill to build now
Build a small routing evaluation rather than selecting a model from product positioning. Choose one difficult repository change, one tool-using workflow and one repeated document or media task. Run each through the relevant candidate model with the same tools, permissions and acceptance checks. Record completion rate, human review time, tool failures, elapsed time and all-in endpoint cost. This gives a team a defensible basis for deciding whether capability tiers actually map to its work.

## Reader takeaway
Google Cloud now lists three Meta Muse models for different kinds of agentic and multimodal work, alongside Muse Spark Open. Treat the advertised distinctions as a shortlist for testing, not proof of production suitability. Validate access, cost, permissions and task-level performance in the deployment configuration your team would actually use.

## Open question
Will Google Cloud publish pricing, quota, regional-access and measured-performance details that let teams compare these models against other Model Garden options on completed tasks?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Google](https://newruntime.com/topics/google.md): Explore the google topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)