K3 API pricing makes cache behavior part of the architecture
The headline prices are $0.30 cached input, $3 cache-miss input, and $15 output per million tokens.
K3 is economical when long, repeated context hits cache. Treat prompt stability and session design as cost controls, not implementation details.
The published rates
Kimi lists K3 at $0.30 per million cached input tokens, $3 per million cache-miss input tokens, and $15 per million output tokens. The context ceiling is 1,048,576 tokens.
A cached input token is priced at one tenth of an uncached input token. Applications that repeatedly send a stable codebase or document corpus should be designed to preserve cacheable prefixes.
What to measure
Track cache-hit input, cache-miss input, output tokens, task completion rate, and retry volume together. Token price alone does not reveal whether a long-running agent is economical.