AWS Neuron 2.31.0 has been released, featuring NKI 0.5.0 with MX FP8 scale dtype support. It also introduces tensor indirection for optimized indexed access patterns. New NkiTensor view APIs enable zero-cost tensor layout transformations. The Neuron UltraServer Operator for Amazon EKS is now in public beta, automating UltraServer management for Trainium workloads. The Neuron Compiler has a redesigned code generation backend, enabled by default on Trn2 and Trn3, improving performance. The Neuron Runtime now supports contiguous shared scratchpad, simplifying device configuration. The NKI Library adds 14 new experimental kernels, including MoE training collectives and deformable attention, with PyTorch reference implementations. Neuron Explorer enhances debugging with System Trace Viewer source code linking and updated default grouping. This release offers significant advancements for AI/ML development on AWS. Users can explore detailed information through the provided documentation links.