AWS AI Updates: April 8, 2026

1. Lambda Response Streaming Now Available in All Commercial AWS Regions

AWS. Lambda’s response streaming capability has expanded to all commercial AWS regions, enabling functions to progressively push response payloads back to clients as data becomes available rather than waiting for full execution to complete. The feature is particularly well-suited for LLM-based applications where streaming tokens reduces perceived latency, and supports payloads up to 200 MB with Node.js and custom runtimes. Previously limited to select regions, full global availability makes this practical for production AI inference workloads at scale. Source

2. Amazon Bedrock Adds Claude Mythos in Gated Research Preview for Cybersecurity

Amazon Bedrock. Anthropic’s Claude Mythos model is now available on Amazon Bedrock as a gated research preview, described as Anthropic’s most advanced model to date with state-of-the-art capabilities in cybersecurity, software coding, and complex reasoning. Access is limited to allow-listed organizations working on defensive cybersecurity applications in the us-east-1 region, part of Anthropic’s Project Glasswing initiative. Interested teams must request access through their Anthropic account executive. Source

3. Uber Expands AWS Partnership to Run Ride-Sharing Workloads on Amazon Trainium2 Chips

AWS. Uber is expanding its AWS infrastructure contract to run ride-sharing compute workloads on Amazon’s custom Trainium2 AI chips, in what TechCrunch describes as a move that sidelines Oracle and Google Cloud. The partnership reflects Amazon’s increasing success in converting large enterprise workloads to its custom silicon, which offers competitive performance-per-dollar for inference-heavy applications. For Uber, Trainium2 applies to real-time demand prediction, driver matching, and pricing models that run continuously at high scale. Source