Open-weight models are shifting how companies build and deploy AI systems.
Kimi K3 from Moonshot AI is now available on Amazon Bedrock. It reaches 2.8 trillion parameters, making it the largest open model of its kind.
The model features a 1-million-token context window and native vision capabilities. It delivers a 2.5x improvement in scaling efficiency compared to Kimi K2.
Key Features
Kimi K3 supports explicit prompt caching to reduce latency and input costs. This helps when reusing context across multiple model calls.
You can invoke the model via cross-Region inference profiles for global workloads. The global profile costs approximately 10% less than geographic profiles.
Security remains strict within AWS data boundaries. Zero data retention is enabled, and no operator access exists during inference.
Why it matters
Runners gain a powerful option for coding and knowledge tasks without changing security posture.
Explicit prompt caching lowers costs for long-running workflows that reuse stable context.
The platform supports tool calling, structured output, and reasoning through standard APIs.
Get started by selecting Kimi K3 in the Amazon Bedrock console playground. You can also call it programmatically using OpenAI-compatible APIs.
Source: https://aws.amazon.com/blogs/machine-learning/introducing-kimi-k3-on-amazon-bedrock/
