Kimi K3.1 Pro Highspeed now available on Amazon Bedrock

Kimi K3.1 Pro Highspeed is part of Moonshot AI's K3.1 series and uses the same 3T-class open-weight architecture as Kimi K3, with substantially more post-training compute. According to Moonshot AI, the extra training targets long-horizon agentic work and fewer factual errors, and it improves financial modeling, office documents, and coding for game engines, 3D scenes and CAD scripts. Moonshot AI reports 2 to 3 times the output speed of Kimi K3.1 Pro.

The model supports tool calling, structured output, configurable reasoning, image input and streaming. AWS says data is processed within the AWS data boundary, is not shared with Moonshot AI or used to train the model, and that zero data retention is always enabled for inference requests.

Developers can call the model through the OpenAI-compatible Chat Completions and Responses APIs as well as the Converse and Invoke APIs, using the global, US or India cross-Region inference profiles. The Flex tier is priced at half the Standard rate, and explicit prompt caching is available on some APIs but not on Converse.

Teams already calling Kimi K3 on Bedrock can test K3.1 Pro Highspeed by changing the model ID, and can set reasoning effort to low, high or max to trade latency against depth.

Agents that resend long system prompts or tool definitions can use explicit cache breakpoints, but the prefix must be at least 256 tokens and cached content is kept for at least 30 minutes.

Cache writes are billed above standard input, so the savings depend on how often the same prefix is reused.