Tech Meridian ← LIVE FEED
PROMY MERIDIAN RU

NEWS · MODELS · #2062

Meta’s Muse Spark Open (816B MoE) added to Amazon SageMaker JumpStart

Meta’s coding-optimized Muse Spark Open (Mixture-of-Experts, 816 billion parameters with 42B active per token and a 1 million-token context window) is now available in Amazon SageMaker JumpStart as two variants (Muse-Spark-Open-816B-A42B BF16 and Muse-Spark-Open-816B-A42B-NVFP4 NVFP4). The model’s open weights (released under the Muse Model License) can be deployed to SageMaker AI endpoints in your AWS account, with instance recommendations ml.p6-b300.48xlarge for BF16 and ml.p5e.48xlarge for NVFP4 and support for the model’s agentic reasoning and multimodal inputs.

KEY POINTS

  1. Meta’s coding-optimized Muse Spark Open (Mixture-of-Experts, 816 billion parameters with 42B active per token and a 1 million-token context window) is now available in Amazon SageMaker JumpStart as two variants (Muse-Spark-Open-816B-A42B BF16 and Muse-Spark-Open-816B-A42B-NVFP4 NVFP4).
  2. The model’s open weights (released under the Muse Model License) can be deployed to SageMaker AI endpoints in your AWS account, with instance recommendations ml.p6-b300.48xlarge for BF16 and ml.p5e.48xlarge for NVFP4 and support for the model’s agentic reasoning and multimodal inputs.
  3. Makes Meta’s large, coding-focused Muse Spark Open model with 1M-token context and open weights easily deployable in customer AWS accounts, lowering friction for enterprise coding and agentic workflows.
MERIDIAN INTELLIGENCE

DECISION BRIEF

75/100
CONFIDENCE
01

WHAT CHANGED

Meta’s Muse Spark Open (Mixture-of-Experts, 816B total parameters, 42B active per token, 1M-token context) is now available in Amazon SageMaker JumpStart as two variants (BF16 and NVFP4). The model’s open weights are released under the Muse Model License. Amazon shows how to deploy the model to a SageMaker AI endpoint in a customer AWS account and returns the model’s reasoning alongside answers. Separately, Meta and Apple publicly signaled collaboration/support and Apple published material about running Muse Spark Open on Mac Studio via the MLX framework.

02

WHY NOW

The model is newly integrated into SageMaker JumpStart, enabling direct deployment in customer AWS accounts and exposing Muse Spark Open’s large context and agentic reasoning capabilities to SageMaker users. Concurrent public collaboration between Meta and Apple (and Apple’s MLX materials) signals increased attention to running the model both in cloud-hosted SageMaker endpoints and on Mac Studio hardware.

03

WHO IS AFFECTED

Directly affected: (1) Amazon SageMaker users and customers who use SageMaker JumpStart and SageMaker AI endpoints; (2) developers and teams working on large codebases, long-horizon or agentic coding workflows (the model was trained on real developer workflows); (3) Mac Studio users and developers following Apple’s MLX work who may run Muse Spark Open locally.

04

CONFIRMED

All confirmed facts are explicitly from the supplied excerpts: - Muse Spark Open is available in Amazon SageMaker JumpStart. (AWS blog) - SageMaker JumpStart offers two variants: BF16 and NVFP4 (Muse-Spark-Open-816B-A42B and Muse-Spark-Open-816B-A42B-NVFP4). (AWS blog) - The model uses a mixture-of-experts architecture with 816 billion total parameters and 42 billion active per token. (AWS blog) - Muse Spark Open supports a 1 million-token context window. (AWS blog) - Meta released the weights under the Muse Model License (including the Muse Spark Usage Policy). (AWS blog) - The model was trained on developer workflows (code generation, debugging, multi-file refactors, long tasks). (AWS blog) - The SageMaker AI endpoint can be deployed in a customer’s AWS account and returns the model’s reasoning alongside answers. (AWS blog) - Meta describes Muse Spark Open as agentic and multimodal (accepts text, images, documents, audio, audio-video and returns text). (AWS blog) - Meta and Apple publicly supported each other’s work on Muse Spark Open, and Apple published technical material about MLX and running Muse Spark Open on Mac Studio. (The Verge)

05

UNCERTAIN

Missing or unclear in the provided excerpts (explicitly labeled): - Recommended SageMaker instance types, cost, latency, and throughput for production deployments are not specified in the supplied excerpts. (missing) - Detailed license terms, permissible uses, and restrictions in the Muse Model License beyond its existence and inclusion of a usage policy are not detailed in the excerpts. (missing) - Exact integration level for tool execution and agent environments on SageMaker (how tools, browsers, and computer-use environments are provided or managed) is described at a high level (“your application or agent framework provides the tools”) but the operational details are not provided. (partial/missing) - Real-world performance, accuracy metrics, or benchmarks for coding, reasoning, or multimodal tasks on SageMaker or Mac Studio are not provided in the excerpts. (missing) - Scope and technical depth of the Meta–Apple collaboration (duration, engineering scope beyond public support and Apple MLX materials) are not specified. (missing) No conflicting evidence was present in the excerpts; information is incomplete on deployment sizing, costs, license constraints, and operational integration.

06

WHAT TO WATCH

Concrete observable signals to watch next (based on the supplied excerpts): - Amazon SageMaker JumpStart model page and JumpStart catalog entries for Muse Spark Open (documentation, deployment guides, example notebooks, SageMaker Studio walkthroughs). (AWS blog mentions Studio and SDK walkthroughs) - SageMaker Python SDK and SageMaker AI developer docs for concrete deployment steps and any recommended instance or autoscaling guidance. (AWS blog shows SDK usage) - Official Muse Model License text and Muse Spark Usage Policy publications or clarifications from Meta. (AWS blog cites the license/policy) - Apple MLX technical post updates, Apple developer documentation, or sample projects showing running Muse Spark Open on Mac Studio. (The Verge cites Apple technical material) - Statements or posts from Meta and Apple leadership or engineering teams that add technical detail on the Mac Studio execution profile or the Meta–Apple collaboration. (The Verge documents CEO/leadership comments) - Community benchmarks, tutorials, or reproducible examples (SageMaker notebooks, GitHub repos, or blog posts) demonstrating model behavior on long-horizon coding tasks or local Mac Studio runs.

GROUNDED IN AWS Machine Learning ↗The Verge AI ↗

WHY IT MATTERS

Makes Meta’s large, coding-focused Muse Spark Open model with 1M-token context and open weights easily deployable in customer AWS accounts, lowering friction for enterprise coding and agentic workflows.

EVIDENCE MAP

4

Editorial claims linked to specific sources, with support, contradiction and context shown separately.

SOURCES & TIMELINE

2
01
AWS MACHINE LEARNING FIRST-PARTY
Meta’s Muse Spark Open now available in Amazon SageMaker JumpStart

Muse Spark Open, Meta’s coding-optimized open-weight model, is now available in Amazon SageMaker JumpStart , the machine learning (ML) hub of Amazon SageMaker AI . The model was trained on real developer workflows, such as multi-file refactors and long debugging sessions, and it supports a context window of 1 million tokens. You can deploy it to a SageMaker AI endpoint in your own AWS account in a few steps, without …

↗
02
THE VERGE AI INDEPENDENT COVERAGE
Meta and Apple team up to make Muse AI models run better on Macs

Meta and Apple’s CEOs are publicly supporting each company’s work on Muse Spark Open. Meta announced the Muse Spark Open AI model today , claiming that it’s the “most capable US open-weight model to date.” But the company also says that it’s the “the first model in its class” that can “run on your desktop using two Apple M5 Ultra Mac Studios with the latest MLX framework.” Meta and Apple, typically rivals , worked …

↗