K3.1 Pro Highspeed is the fast lane for Kimi's flagship model: the same K3.1 Pro answers, streamed about 2–3× faster by Moonshot's figure, for twice the price per token. Here's how fast it really is, what it costs and when it's worth it.
Launched with K3.1 Pro on 23 October 2026 in the Kimi apps, Kimi Work, Kimi Code and the Kimi API.
Moonshot's figure. Independent checks measured about 1.7–2×.
Input and output, double standard K3.1 Pro.
The same as standard Pro, so cached prompts cost no more.
Same model, same abilities, same context window.
| Feature | K3.1 Pro | K3.1 Pro Highspeed |
|---|---|---|
| API model ID | kimi-k3.1-pro | kimi-k3.1-pro-highspeed |
| Kimi Code model ID | k3.1-pro | k3.1-pro-highspeed |
| Model and abilities | K3.1 Pro | Same model and abilities |
| Output speed (Moonshot) | Standard | About 2–3× faster |
| Output speed (independent checks) | ≈ 57–88 tokens/s | ≈ 113–150 tokens/s |
| Input, per 1M tokens | $2.00 | $4.00 |
| Cached input, per 1M tokens | $0.20 | $0.20 (same) |
| Output, per 1M tokens | $10.00 | $20.00 |
| Cache write, 5 min / 1 hour | $2.00 / $4.00 | $4.00 / $8.00 |
| Context window | 1M tokens | 1M tokens |
| Thinking | Always on: low, high or max effort | Same |
| Kimi app access | All plans (limited on Free) | Pro plan ($39/month) and up |
| Kimi Code quota | Standard | About 2× usage per task, Pro plan and up |
Prices from the Kimi API pricing page, per million tokens, before tax. Full model details: Kimi K3.1 Pro guide.
Moonshot says Highspeed runs at about 2–3× the speed of standard K3.1 Pro. Independent checks in October 2026 found a smaller but still large gain.
Speeds change with server load, prompt length and thinking effort. Treat these as a guide, not a guarantee.
Measured output speed from two independent checks.
Enter a typical request. See what Highspeed adds to your bill and how much waiting it saves.
Assumes 70 tokens/s for standard Pro and excludes cache-write fees and thinking time before the first token. Estimates only.
Interactive coding in Kimi Code, live chat features in your app, demos and pair-programming, where every reply holds you up.
Agents that make dozens of model calls in a row add up the wait each time. Faster output shortens the whole run.
Overnight jobs, bulk document processing and anything nobody watches. Speed doesn't matter, and Pro is half the price.
Highspeed doubles input and output prices. In Kimi Code it uses about twice the quota, so your plan runs out sooner.
Tool calls, web searches, test runs and scripts take the same time. If an agent spends most of its time running commands, the overall gain is smaller.
K3.1 Pro always thinks before it answers. Highspeed streams faster, but you may still wait a while before the first word appears. Lower the effort for quicker starts.
Highspeed has the same abilities as K3.1 Pro. Pick it for speed, not for better answers.
Highspeed is Moonshot's faster serving of K3.1 Pro. The open weights are for K3.1 Pro; speed when you self-host depends on your own hardware.
On the Pro plan or higher, pick K3.1 Pro Highspeed in the model menu. Kimi Work needs version 3.1.0 or later.
On the Pro plan or higher, type /model and choose K3.1 Pro Highspeed. It uses about twice the quota.
Set the model to kimi-k3.1-pro-highspeed. Nothing else changes from standard K3.1 Pro.
Keys, billing and rate limits: Kimi API guide. Coding setup: Kimi Code guide. Plan prices: Kimi pricing.