Introducing Kimi K3.1 Pro on Amazon Bedrock

AWS Blog
Read full story· 8 min
Moonshot AI has launched Kimi K3.1 Pro Highspeed, a faster, post-trained version of its 3-trillion parameter Kimi K3 model, now accessible via Amazon Bedrock. The model offers enhanced long-context handling, improved accuracy, and supports advanced tasks like financial modeling and coding, with secure data handling on AWS.

What happened

Moonshot AI has released Kimi K3.1 Pro Highspeed, a new version of its Kimi K3 model, on Amazon Bedrock. This updated model is designed to deliver faster performance and improved accuracy for users. It is available to customers through the same APIs and endpoints as earlier Kimi K3 models on AWS. The launch brings enhanced capabilities for handling complex tasks and ensures data is processed securely within AWS infrastructure.

Key details

  • Kimi K3.1 Pro Highspeed is a post-trained, 3-trillion parameter model with a 1-million-token context window.
  • The model generates output tokens 2–3 times faster than Kimi K3.1 Pro.
  • It supports advanced tasks like financial modeling, office documents, and coding for game engines and CAD scripts.
  • Data is processed within AWS boundaries, with zero data retention and no operator access to user prompts or completions.
  • A Flex tier is available, offering lower costs in exchange for potentially slower response times.

Why it matters

Businesses and developers using Amazon Bedrock can now access a faster and more capable AI model for demanding tasks such as financial analysis and software development. Organizations with strict data privacy needs benefit from processing that keeps all data within AWS and prevents sharing with the model provider. The Flex tier allows customers to manage costs more effectively by choosing lower prices for workloads that can tolerate longer response times.


More in LLM & Text Generation