Kimi K3 Model Launches on Amazon Bedrock
Moonshot AI's 2.8‑trillion‑parameter Kimi K3 is now available on Amazon Bedrock, bringing vision, a 1‑million‑token context window, and explicit prompt caching for long‑running coding and knowledge workflows.
Original source published: September 18, 2026
Kimi K3, the latest open‑weight model from Moonshot AI, is now offered on Amazon Bedrock. The model features 2.8 trillion parameters, native vision capabilities, a 1‑million‑token context window, and an approximate 2.5× scaling‑efficiency improvement over its predecessor Kimi K2. It is also the first open‑weight model on Bedrock to support explicit prompt caching, which can lower latency and input‑token costs when reusing stable context.
Bedrock users can access Kimi K3 through the console, OpenAI‑compatible APIs, or cross‑Region inference profiles (global.moonshotai.kimi‑k3 and us.moonshotai.kimi‑k3). The service maintains data within the AWS boundary, enforces zero data retention for inference, and blocks operator access to prompts and completions, preserving data confidentiality.
For learners and developers, Kimi K3 demonstrates how open‑weight models can be integrated into production‑grade platforms while offering large context windows and caching mechanisms that support efficient, long‑horizon coding and research tasks.