Kimi K3 Model With 2.8 Trillion Parameters Arrives on Amazon Bedrock

01 What happened

Moonshot AI has integrated its Kimi K3 model into Amazon Bedrock. The company states that this is the first open model to reach 2.8 trillion parameters.

02 Key details

• The model supports a 1-million-token context window and includes native vision processing capabilities.
• It is the first open-weight model on Amazon Bedrock to support explicit prompt caching, which reduces input latency and costs.
• Moonshot AI reports that the architecture provides a 2.5x improvement in scaling efficiency compared to the Kimi K2 model.

03 Why it matters

This release enables developers to manage long-context workflows more efficiently by utilizing prompt caching to optimize performance and costs during complex coding and data analysis tasks.

04 Who it matters to

Software developers, machine learning engineers and IT architects.