Build planning from official docs
Translate the Kimi API and Kimi Code entries into a concrete build checklist: endpoint shape, reasoning effort, context settings, token cost planning and current caveats.
This short video summarizes the page: the source set, the decision boundary and the practical next action.
The source ledger records the official quickstart as OpenAI-compatible, with base_url set to https://api.moonshot.ai/v1 and example model kimi-k3. Kimi Code uses separate coding endpoints and model configuration.
The K3 API guide notes that K3 always thinks. Developers can adjust reasoning_effort across low, high and max, and can set max_completion_tokens up to 1,048,576 when a workflow needs the long context ceiling.
At collection time, the API platform showed cache hit $0.30/MTok, input $3.00/MTok and output $15.00/MTok. Treat those as planning defaults and verify the current platform page before billing decisions.
Defaults come from the source ledger's API platform snapshot. Recheck the official platform before production billing.
Cards are generated from the clauxel/kimik3 source ledger and keep source type, priority and status visible.
Official API platform entry listing Kimi K3 as the flagship model and showing current public price points at collection time.
Official quickstart with OpenAI-compatible API shape, base URL and a Kimi K3 example model.
Official K3 API guide covering reasoning effort, streaming, vision input, structured output, tool choice, dynamic tools and 1M context limits.
Official K3 pricing reference for pay-as-you-go billing, cache hit, input and output metering.
Official Kimi Code model configuration page with model IDs, context configuration and compatible coding endpoints.
Official changelog showing the K3 release window and model-list refresh behavior for API-key users.