AWS launches Moonshot AI’s Kimi K3 on Amazon Bedrock
Summary
AWS launches Moonshot AI’s Kimi K3 on Amazon Bedrock via U.S. and global cross-Region inference, bringing a 2.8-trillion-parameter open-weight model with native vision, a 1-million-token context window, and Bedrock’s first explicit prompt caching.
Key Points
- AWS launches Moonshot AI's Kimi K3 on Amazon Bedrock through US and global cross-Region inference profiles.
- Moonshot AI says Kimi K3 is its most capable model, with 2.8 trillion parameters, native vision, a 1-million-token context window, and roughly 2.5x Kimi K2's scaling efficiency.
- Kimi K3 becomes Bedrock's first open-weight model with explicit prompt caching; cached prompt prefixes require at least 1,024 tokens and remain stored for at least 30 minutes.