How to cut your Moonshot Kimi API bill

Mode

Automatic: repeated prefixes are cached with no code change

Automatic prefix caching (prompt_cache_options.mode supports only implicit); TTL tier selectable. On the Anthropic-compatible Messages API, a top-level cache_control sets the TTL.

Cache lifetime

5 minutes by default or 1 hour option; each hit renews the entry under its original (locked) TTL at no extra cost

Minimum prompt

Not stated; cache stored in blocks, and a portion smaller than one full block is a cache miss

Cache write cost

kimi-k3: cache writes billed separately, 5m write at the input price, 1h write at 2x; K2 series (kimi-k2.7, k2.7-highspeed, k2.6) have no cache-write charge

Derived from the pricing table. Computed from the pricing table: kimi-k3 Cache Write (5min) $3.00 equals the $3.00 Input Price; Cache Write (1h) $6.00 is 2x. The K2 no-write-charge part is quoted.

Cache read cost

kimi-k3 cache hit is one-tenth of the cache-miss price

Storage cost

Not stated

How to enable

Automatic. Choose TTL with prompt_cache_options (Chat Completions/Responses) or top-level cache_control (Messages).

Request fragment

Copy

"prompt_cache_options": {"mode": "implicit", "ttl": "1h"}

From the docs

“When `prompt_cache_options` is omitted, the system uses the `5m` TTL by default: prefixes that meet the hit conditions are automatically written to the cache and reused, and Cache Write charges apply.”

“Input that hits the cache within its lifetime is billed only at the Cached Input Price: each hit automatically refreshes the cache entry's lifetime, with no additional cache write charge.”

“`prompt_cache_options.mode` currently supports only `implicit`, and `ttl` supports `5m` and `1h`.”

“Cache is stored in blocks. A portion smaller than one full block cannot be written to cache; it counts as a cache miss and is billed at the regular input (cache-miss) rate.”

“| Cache Write (`5m`) | $3.00 | Charged once when the prefix is first written; valid for 5 minutes, and each hit resets the validity to 5 minutes |”

“| Cache Write (`1h`) | $6.00 | Charged once when the prefix is first written; valid for 1 hour, and each hit resets the validity to 1 hour |”

“`kimi-k3` supports Cache Write; `kimi-k2.7`, `kimi-k2.7-highspeed`, and `kimi-k2.6` do not.”

“for `kimi-k3`, the cache-hit price is only one-tenth of the cache-miss price”

“A cache entry's TTL is locked at first write and cannot be changed later”

Sources: platform.kimi.ai/docs/pricing/chat, platform.kimi.ai/docs/guide/context-caching

Discount

40% off select models

40% off (Batch costs 60% of the standard model price)

Window

completion_window 12h minimum to 7d maximum (e.g. 12h, 1d, 3d); tasks not finished in the window expire

Limits

Input file non-empty, max 100MB; up to 1000 batch-purpose files per organization; not subject to real-time concurrency limits; supports kimi-k2.7-code and kimi-k2.6 only

Endpoint

POST /v1/batches

Stacks with caching

Not stated: the provider's pages do not say

From the docs

“Batch API inference costs are 60% of the standard model price, ideal for large-scale tasks with low real-time requirements.”

“It supports submitting tasks in bulk via files, saving 40% on inference costs compared to real-time API calls.”

“Time window for task processing, supports formats like 12h, 1d, 3d, minimum 12h, maximum 7d”

“Tasks must complete within the specified `completion_window`, otherwise they expire”

“| File size | Must be non-empty, max 100MB |”

“Batch API supports `kimi-k2.7-code` and `kimi-k2.6` models”

Sources: platform.kimi.ai/docs/pricing/batch.md, platform.kimi.ai/docs/guide/use-batch-api.md, platform.kimi.ai/docs/api/batch-create.md

Other discounts on the pages checked: Not stated