# API: Kimi K3 always thinks; thinking.type is not supported.
# Effort: reasoning_effort = "low" | "high" | "max" (default "max").
# https://platform.kimi.ai/docs/api/models-overview
# Pricing: https://platform.kimi.ai/docs/pricing/chat-k3 (accessed 2026-07-17).
# Cache write (same page, accessed 2026-10-07): $3.00 for the default 5min TTL,
# $6.00 for the 1h TTL; cache_write uses the default 5min price.
# Chat API max_completion_tokens can be set up to 1048576 for Kimi K3:
# https://platform.kimi.ai/docs/api/chat
base_model = "moonshotai/kimi-k3"

[[reasoning_options]]
type = "effort"
values = ["low", "high", "max"]

[interleaved]
field = "reasoning_content"

[cost]
input = 3.0
output = 15.0
cache_read = 0.3
cache_write = 3.0

[limit]
output = 1_048_576