Kimi Code offers two primary models: Kimi K3 and Kimi K2.7 Code, accessible through four model IDs: `k3`, `k3-256k`, `kimi-for-coding`, and `kimi-for-coding-highspeed`. The `k3` model boasts a 1M context window, while `k3-256k` is a more cost-efficient version with a 256k context window. `kimi-for-coding` and `kimi-for-coding-highspeed` are based on Kimi K2.7 Code, with the latter offering significantly faster output.

Switching between models, especially from the 1M context `k3` to `k3-256k`, requires careful consideration. If the current session exceeds 256k tokens, manual context compaction is recommended to avoid data loss or session failure, particularly if video inputs were used, as `k3-256k` does not support video. When switching from `k3-256k` to `k3` (1M), context cache is not affected, allowing for direct switching.

Membership tiers influence model access and context window sizes. For instance, 1M context for `k3` is available on higher-tier plans. Users experiencing `401` errors should verify their plan entitlements for specific models and context limits. Starting new sessions after model switches is advised for optimal performance and reduced token consumption.