NewModelHubCalculate Your Cost
Enter uncached, cached-hit, and cache-write tokens separately. If your usage report gives total input plus a cached subset, subtract cached from total before filling uncached input. Cache-write tokens are the stored prefix size, not the hit count.
Your Estimate
Cost per request: $0.051600
Estimated monthly cost: $5,160.00
Kimi K3.1 Pro Rate Card (per 1M tokens)
| Model | Input (uncached) | Output | Cached input | Cache write 5m | Cache write 1h |
|---|---|---|---|---|---|
| kimi-k3.1-pro (Standard) | $2.00 | $10.00 | $0.20 | $2.00 | $4.00 |
| kimi-k3.1-pro-highspeed (Highspeed) | $4.00 | $20.00 | $0.20 | $4.00 | $8.00 |
Rates from Moonshot's official pricing page (platform.kimi.ai/docs/pricing/chat), verified October 25, 2026 (Asia/Shanghai). Model IDs: kimi-k3.1-pro, kimi-k3.1-pro-highspeed. The official table lists a single rate per column — no length-based tiers are shown — across a 1,048,576-token context window. Pay-as-you-go, no subscription; prices exclude applicable taxes. Caching is enabled automatically for repeated request prefixes (default TTL 5 minutes); a cache hit bills only the cached-input rate. Minimum top-up $1 to start; a $5 voucher is granted at $5 cumulative top-up (voucher excluded from the cumulative total).
What Drives Your Bill
- Output dominates, and thinking bills as output. At $10.00 vs $2.00 per million, output length is the main lever — and thinking is always on, so effort max (the default) spends the most reasoning tokens. Drop effort to low or high for routine turns.
- Cache hits are 10× cheaper than fresh input. Cached input is $0.20 vs $2.00 fresh. Put stable system prompts, tool schemas, and reused context in a repeated prefix so Kimi can hit cache.
- Cache writes are not free. Storing a prefix bills $2.00 (5m) or $4.00 (1h) per million tokens in Standard — double in Highspeed. Match TTL to your real reuse horizon; a 1-hour TTL on a prefix you only reuse for minutes wastes write fees.
- Highspeed is a latency premium. It doubles input and output prices while cache-hit price stays $0.20. Worth it for interactive coding loops; for overnight batches, Standard finishes the same work for half the token price.
- Do not double-count input. Uncached + cached must partition the billed input, and cache-write tokens are a third, separate bucket — pasting total input into every field inflates the estimate.
Related tools
FAQ
How much does Kimi K3.1 Pro cost?
Official Moonshot API list price is $2.00 per million input tokens, $10.00 per million output tokens, and $0.20 per million cached-input (cache-hit) tokens. Cache writes cost $2.00 per million for a 5-minute TTL and $4.00 per million for a 1-hour TTL.
What is the difference between Standard and Highspeed?
Highspeed is about 2-3 times faster at twice the input and output price ($4.00 input, $20.00 output per million). Cached-input price stays $0.20 per million in both tiers, while cache-write fees double ($4.00 for 5 minutes, $8.00 for 1 hour).
How does context caching billing work?
Caching is enabled automatically for repeated request prefixes, with a default TTL of 5 minutes. Storing a prefix bills the cache-write rate for the chosen TTL; a later cache hit bills only the $0.20 cached-input rate. Enter cached-hit tokens and cache-write tokens separately, and do not include cached tokens in the uncached input field.
What are the model IDs and thinking settings?
Use kimi-k3.1-pro for Standard and kimi-k3.1-pro-highspeed for Highspeed. Thinking mode is always on, with reasoning effort low, high, or max (default max). Context window is 1,048,576 tokens.
Is this an official Moonshot tool?
No. It is an unofficial, fan-made estimator. Moonshot's invoice and platform.kimi.ai pricing are authoritative. This calculator covers cloud API rates only; model weights are promised by November 4, 2026, so self-hosting costs are not modeled here.
Methodology & sources: rates verified against Moonshot's official pricing page on October 25, 2026 (Asia/Shanghai) — https://platform.kimi.ai/docs/pricing/chat. Estimates exclude taxes, reseller markups, and non-API products. Your invoice follows metered usage.