Introducing Kimi K3 on Amazon Bedrock
Kimi K3 from Moonshot AI is now available on Amazon Bedrock, giving you a powerful new open-weight option for coding and knowledge work. It offers native vision, a 1-million-token context window, and explicit prompt caching to reduce latency and input costs.
Open-weight models are revolutionizing the AI industry by providing a flexible, scalable solution for building and deploying intelligent systems. AWS has committed to supporting this trend, adding a range of open-weight models to Amazon Bedrock, their cloud-based service for generative AI. One such model, Kimi K3 from Moonshot AI, has recently been made available on Bedrock.
Kimi K3 stands out with its impressive capabilities, boasting 2.8 trillion parameters and native vision capabilities. It also features a large 1-million-token context window, enabling it to handle complex, multi-step tasks such as coding and knowledge work. Notably, Kimi K3 offers a 2.5x improvement in scaling efficiency compared to its predecessor, Kimi K2.
A key feature of Kimi K3 is its support for explicit prompt caching, which can significantly reduce latency and input costs when working with large repositories or documents. This allows for the reuse of context across model calls, making it an ideal choice for long-running workflows that require sustained context.
Moonshot AI's decision to bring Kimi K3 to Amazon Bedrock reflects AWS's ongoing investment in open-weight models. Since 2025, Bedrock has integrated dozens of open-weight models from leading providers like DeepSeek, Google, NVIDIA, and OpenAI. As new models become available, Bedrock continues to enhance its inference technology to ensure optimal performance.
Using Kimi K3 on Amazon BedRock is straightforward. Users can simply access the Amazon Bedrock console, go to the Test Playground, and select Kimi K3 as the model. For programmatic access, the bedrock-runtime endpoint supports the OpenAI-compatible Responses and Chat Completions APIs, along with the Amazon Bedrock Invoke and Converse API APIs. The global.moonshotai.kimi-k3 profile can be used for cross-Region inference, offering cost savings compared to geographic profiles.
To optimize inference with explicit prompt caching, users can add a prompt_cache_breakpoint to the input content after at least 1,024 tokens. This enables Amazon Bedrock to reduce response latency and input token costs for matching subsequent requests. The OpenAI Python API and the aws-bedrock-token-generator library provide a convenient way to generate short-term bearer tokens for authentication to Amazon Bedrock.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
