Industry
Deploying Kimi K3 on AWS
Kimi K3 is a 2.8‑trillion‑parameter Mixture of Experts model released by Moonshot AI on July 27, 2026, with its weights publicly available for self-hosting. Its architecture—using Kimi Delta Attention
Kimi K3 is a 2.8‑trillion‑parameter Mixture of Experts model released by Moonshot AI on July 27, 2026, with its weights publicly available for self-hosting. Its architecture—using Kimi Delta Attention, Gated Multi‑Head Latent Attention, and Stable LatentMoE—spreads 2.8 T parameters across 896 experts (activating only 16 per token), enabling ~104 B active parameters per forward pass. The model can be deployed on AWS via Amazon SageMaker HyperPod or a fully managed EKS cluster for scalable inference of advanced reasoning, multi‑step agentic workflows, and long‑horizon coding tasks.
Related
- Monitoring discriminative ML models using Amazon SageMaker AI with MLflow
- As always, Cosmos 3 Edge is fully open. This includes model weights, post-training recipes and code. Available now on @huggingface 🤗 https:…
- Command A+ from @cohere is out now :) its our best model yet and its open source apache 2.0
- Monitor Amazon SageMaker Pipelines cross-account with custom Amazon CloudWatch dashboards
Source: AWS ML Blog | 2026-07-30