name = "Moonshot AI"
env = ["MOONSHOT_API_KEY"]
npm = "@ai-sdk/openai-compatible"
# Chat is POST `/v1/chat/completions`. Kimi K2.5 and K2.6 toggle reasoning with
# `thinking.type = "enabled"|"disabled"` (enabled by default); K2.7 Code only
# accepts `"enabled"`, so it is always on. K3 also always thinks, but exposes
# `reasoning_effort = low|high|max` instead of the K2.x `thinking` field.
# https://platform.kimi.ai/docs/api/chat (accessed 2026-06-25)
doc = "https://platform.moonshot.ai/docs/api/chat"
api = "https://api.moonshot.ai/v1"