Meta’s Muse Spark Open now available in Amazon SageMaker JumpStart
aws ·
Meta’s Muse Spark Open, a coding-optimized open-weight model, is now available in Amazon SageMaker JumpStart. Trained on real developer workflows, the model features a mixture-of-experts architecture with 816 billion total parameters and supports a context window of one million tokens. It offers two variants in SageMaker: BF16 and NVFP4 quantized, which differ significantly in memory requirements and instance selection. The NVFP4 variant runs on ml.p5e.48xlarge instances, while the BF16 variant requires ml.p6-b300.48xlarge instances due to higher memory needs. Users can deploy the model via SageMaker Studio or the SageMaker Python SDK, creating endpoints within their own AWS accounts for full control over security and infrastructure. As a reasoning model, it provides both final answers and detailed reasoning content, making it suitable for complex agentic coding tasks. Deployment involves accepting Meta's license agreement and configuring specific compute resources based on the chosen variant.
aws