Try Kimi at kimi.com →

Kimi K2.7 Code Highspeed
Up to 260 Tokens a Second

K2.7 Code Highspeed is the fastest way to run Kimi's June 2026 coding model: about 180 tokens per second on typical coding tasks and up to 260 on short ones. It's still in Kimi Code, but the API version retires by the end of December 2026. Here's what it does, what it costs and what to move to.

~180–260 tokens/s $1.90 / $8 per 1M 256K context
0 50 100 150 200 250 300 TOKENS PER SECOND ~180typical coding task 260short-context peak ~180typical coding task Up to 6× standard K2.7 Code api/routes.ts 1 export 2 async function 3 getOrders(req) { 4 const user = 5 await auth(req); 6 const rows = 7 await db.orders 8 .where({ user }) 9 .limit(50); 10 return 11 json(rows); 12 } ✓ Done. Run tests?
STATUS

Status: still in Kimi Code, retiring from the API

Kimi API: closing

Closed to new users since 31 October 2026. Existing users can keep using it, but Moonshot plans to retire all K2-series models by the end of December 2026.

Kimi Code: available

Still listed as kimi-for-coding-highspeed on the Pro plan and above, as of November 2026.

Moonshot's replacement

Moonshot points API users to Kimi K3.1 Pro. For speed, use K3.1 Pro Highspeed.

OVERVIEW

K2.7 Code Highspeed at a glance

~180Tokens per second

On typical coding tasks, rising to 260 on short ones (Moonshot).

6×Faster

Up to six times standard K2.7 Code, by Moonshot's figure.

256KToken context

Text, image and video input, with thinking always on.

$1.90 / $8API per 1M tokens

Input and output, double standard K2.7 Code.

Launched 15 June 2026, three days after K2.7 Code, first for Kimi Code beta members, API developers and Kimi Business users, with limited capacity at the start.
COMPARE

K2.7 Code vs Highspeed vs K3.1 Pro Highspeed

FeatureK2.7 CodeK2.7 Code HighspeedK3.1 Pro Highspeed
API model IDkimi-k2.7-codekimi-k2.7-code-highspeedkimi-k3.1-pro-highspeed
Kimi Code model ID(replaced by K2.8 Preview)kimi-for-coding-highspeedk3.1-pro-highspeed
Output speedStandard~180 tokens/s, up to 260 (Moonshot)~113–150 tokens/s (measured)
Context window256K tokens256K tokens1M tokens
ThinkingAlways onAlways on, no effort settingAlways on: low, high or max
API input / output$0.95 / $4.00$1.90 / $8.00$4.00 / $20.00
Cached input$0.19$0.38$0.20
Kimi Code quota–About 3× regular modelsAbout 2× K3.1 Pro
Kimi Code plan–Pro and upPro and up
API statusClosed to new users; retiring by end of Dec 2026Closed to new users; retiring by end of Dec 2026Current

API prices per million tokens from Moonshot's pricing page. Kimi Code details from its model table, November 2026.

SPEED

How fast is K2.7 Code Highspeed?

It's the fastest model Kimi offers. A typical 2,000-token answer streams in about 11 seconds at 180 tokens per second.

  • 1
    Moonshot's figures: about 180 tokens/s on coding tasks with median-length inputs and up to 260 tokens/s on short-context tasks.
  • 2
    Compared with K3.1 Pro Highspeed: independent tests measured that at about 113–150 tokens/s, so K2.7 Code Highspeed is still quicker, but it's a much older and weaker model.
  • 3
    Thinking still takes time: it always thinks before answering and has no effort setting, so very short tasks don't get much faster to start.
OUTPUT SPEED, TOKENS/SFastest Kimi option, but an older model K2.7 Code Highspeedshort context · Moonshot 260 K2.7 Code Highspeedtypical coding · Moonshot ~180 K3.1 Pro Highspeedmeasured · Independent 113–150 K3.1 Promeasured · Independent 57–88 Moonshot figures and independent tests aren’t measured the same way K2.7 Code Highspeed: Moonshot’s launch figures, June 2026. K3.1 Pro: AIHubMix and Artificial Analysis data, October 2026.

Output speed of Kimi's fast options.

WHEN

When to use it, and when not to

Fast, simple edits

Renames, small fixes, boilerplate and tests where you want the answer in seconds and the task isn't hard.

Tight edit-run loops

When you iterate quickly, a fast model that can fix its own mistakes may beat a slower, smarter one.

Not for the hardest work

For big refactors or tricky bugs, K3.1 Pro (or its Highspeed version) is much stronger and has a 1M context.

Not for long-term API projects

The API version retires by the end of December 2026. Build new integrations on K3.1 Pro.

USE

How to use K2.7 Code Highspeed

Kimi Code

On the Pro plan or higher, type /model and pick kimi-for-coding-highspeed. Each task uses about 3× the quota of regular models.

Kimi API (existing users)

Use kimi-k2.7-code-highspeed at https://api.moonshot.ai/v1 until the K2 series retires.

Other coding tools

Connect with a Kimi Code key to https://api.kimi.ai/coding/v1 and the model kimi-for-coding-highspeed.

# Kimi Code (Pro plan and up) /model # choose kimi-for-coding-highspeed → K2.7 Code Highspeed # Other tools with a Kimi Code key base_url: https://api.kimi.ai/coding/v1 model: kimi-for-coding-highspeed
MIGRATE

Moving to K3.1 Pro before December

  • 1
    Change the model ID. The API is the same, so switching is usually a one-line change.
  • 2
    Expect higher prices. K3.1 Pro Highspeed costs $4/$20 per million tokens against $1.90/$8, though it's a much stronger model.
  • 3
    Use the effort setting. K3.1 Pro lets you set low thinking effort for quick tasks, which K2.7 Code didn't offer.
  • 4
    Test before the cut-off. Run your usual prompts on K3.1 Pro now so there are no surprises when K2 retires.

Full API setup: Kimi API guide. Speed and cost trade-offs: K3.1 Pro Highspeed.

Migration
# Before (retiring by end of December 2026) model="kimi-k2.7-code-highspeed" # After: same speed-first role, stronger model, 1M context model="kimi-k3.1-pro-highspeed" reasoning_effort="low" # fastest start; use "high" for harder tasks # Or, to keep costs down model="kimi-k3.1-pro"
K2.7

About Kimi K2.7 Code

A coding specialist

Released on 12 June 2026, K2.7 Code is a thinking-only coding model built for long tasks. Moonshot reported better success on runs of 4,000+ tool calls and 12+ hours.

Less overthinking

Moonshot says it uses about 30% fewer thinking tokens than K2.6 on average, which cuts cost and waiting on simple jobs.

Open weights

The weights are on Hugging Face at moonshotai/Kimi-K2.7-Code under a Modified MIT licence, so it stays usable after the API retires if you host it yourself.

K2.7 Code was replaced behind Kimi Code's default ID by K2.8 Preview on 11 September 2026. See all versions on the Kimi models page.

FAQ

K2.7 Code Highspeed FAQ

Kimi K2.7 Code Highspeed is a faster way of running Kimi K2.7 Code, Moonshot's June 2026 coding model. It's the same model with the same abilities, served at about 180 tokens per second on typical coding tasks and up to 260 on short-context tasks.

Moonshot says about 180 tokens per second on coding tasks with median-length inputs and up to 260 tokens per second on shorter tasks, up to 6 times faster than standard K2.7 Code.

On the Kimi API it costs $1.90 per million input tokens, $0.38 per million cached input tokens and $8 per million output tokens, twice the standard K2.7 Code price. In Kimi Code it uses about 3 times the quota of regular-speed models.

On the Kimi API, yes. K2-series models, including K2.7 Code and K2.7 Code Highspeed, were closed to new users on 31 October 2026, and Moonshot plans to retire them all by the end of December 2026. Existing users can keep using them until then.

Moonshot points API users to Kimi K3.1 Pro. If speed matters most, use K3.1 Pro Highspeed. It's stronger, has a 1M-token context and is the current fast option.

Yes. As of November 2026, Kimi Code still lists it as kimi-for-coding-highspeed, on the Pro plan and above, with a 256K context and about 3 times the quota of regular models.

K2.7 Code has open weights on Hugging Face (moonshotai/Kimi-K2.7-Code) under a Modified MIT licence. Highspeed is Moonshot's fast hosted version; self-hosted speed depends on your own hardware.

No. kimi-for-coding now runs K2.8 Preview (since 11 September 2026). kimi-for-coding-highspeed still runs K2.7 Code Highspeed.

Need speed after December?

K3.1 Pro Highspeed is Kimi's current fast model, with a 1M context and far stronger coding.