pub fn kimi_code_capability(model_id: &str) -> Option<KimiReasoningCapability>Expand description
Reasoning capability for an exact managed Kimi Code model id.
Evidence: the managed /models payload provisioned by the official CLI —
k3/k3-256k carry think_efforts {low, high, max} with default high
and always_thinking; the kimi-for-coding* pair are always_thinking
with no effort control. Reasoning is not disableable on any managed id.