Moonshot puts Kimi K3 on Amazon Bedrock with a million‑token window
Moonshot describes it as the first open model at 2.8 trillion parameters. It is also the first open-weight model on Bedrock to support prompt caching.
Kimi K3, an open-weight model from the Chinese company Moonshot AI, became generally available on Amazon Bedrock on Thursday. Moonshot describes it as the first open model to reach 2.8 trillion parameters.
The model combines native vision with a context window of one million tokens. Moonshot puts the scaling efficiency at roughly 2.5 times that of Kimi K2, a figure from the company rather than from an independent evaluation.
AWS says it is the first open-weight model on Bedrock to support explicit prompt caching, which lets a caller reuse context across requests instead of resending it. For long agent runs and repeated work over the same repository, that changes the cost profile rather than the capability.
The uses AWS points to are long coding sessions across large codebases, analysis spanning many documents including scanned pages, and extended agent workflows. All three are cases where the context window is the binding constraint.
Availability is a matter of record on AWS. The performance claims are Moonshot's own and have not been independently reproduced.