Introducing Reflection Beam: an open-weight model for agentic coding on AWS
Machine Learning Blog
This article introduces Reflection Beam, an open-weight sparse mixture-of-experts model designed for agentic coding and multi-step software engineering tasks, now available on AWS.
- Beam has 501 billion total parameters and 23 billion active parameters with 1 million token context length
- Deploy via SageMaker JumpStart for fully managed endpoints or HyperPod Inference for Kubernetes-native control
- Available in BF16, FP8, and NVFP4 checkpoint variants with different memory footprints and hardware compatibility
- Supports SGLang and vLLM runtimes for custom inference configurations and distributed execution
- Model caching and intelligent routing optimize performance across multi-GPU deployments
- Available in US East (N. Virginia) and US East (Ohio) regions
Beam enables developers to build software engineering and agentic systems with open-weight model control over deployment, runtime, and customization on AWS infrastructure.
The AWS News Feed is currently looking for gold sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.
Related articles
Nov 17
2026
2026
AWS announces agentic account creation for agentic coding harnesses
Nov 17
2026
2026
Building a gated review pipeline for AI coding agents on AWS
Oct 20
2026
2026
Build a semantic layer for agentic AI on AWS with AtScale and Amazon Bedrock AgentCore
Oct 20
2026
2026
Build a semantic layer for agentic AI on AWS with Strategy Mosaic and Amazon Bedrock AgentCore
The AWS News Feed is currently looking for silver sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.