Kimi K3 by Moonshot AI is now generally available on Amazon Bedrock
Moonshot AI’s Kimi K3, a 2.8-trillion-parameter open model with vision and a 1-million-token context window, is now generally available on Amazon Bedrock for coding and knowledge tasks.
Amazon Bedrock has added Kimi K3 from Moonshot AI as a generally available model, expanding the platform’s open-weight portfolio while maintaining the same security and governance standards customers expect. The model is positioned as Moonshot AI’s most capable offering and is described as the first open model to reach 2.8 trillion parameters. It integrates native vision capabilities alongside a 1-million-token context window, enabling long coding sessions, multi-document analysis, and extended agent workflows. On Amazon Bedrock, Kimi K3 operates within the same security boundary as proprietary models, ensuring consistent access controls, encryption, and auditing across deployments.
Moonshot AI states that Kimi K3 delivers approximately 2.5 times better scaling efficiency compared to its predecessor, Kimi K2, reflecting improvements in model architecture and training efficiency. The model supports explicit prompt caching on Amazon Bedrock, a feature not previously available for open-weight models on the platform, which reduces latency and lowers input costs when reusing context across multiple model calls. This capability is designed to improve efficiency for repetitive or iterative workflows, such as debugging or reviewing large codebases.
Kimi K3 is the first open-weight model on Amazon Bedrock to offer explicit prompt caching, a feature that helps minimize latency and reduce input costs during repeated context usage. The model is accessible in all AWS Regions where Amazon Bedrock is available, with cross-Region inferencing enabled to support global deployments. Customers can begin using Kimi K3 immediately through the Amazon Bedrock Console, with additional guidance available in the platform’s documentation and a dedicated launch blog post.
The addition of Kimi K3 follows Amazon Bedrock’s strategy of broadening its open-weight model offerings while preserving security and governance controls. By integrating Kimi K3, customers gain access to a high-parameter open model optimized for coding and knowledge-intensive tasks, supported by Amazon Bedrock’s existing infrastructure for access management, encryption, and auditing. The model’s combination of large context windows, vision capabilities, and prompt caching aims to enhance productivity for developers and analysts working with complex or multi-modal data.