Try Kimi at kimi.com →

Kimi K2.8 Preview
Kimi Code's Everyday Model

K2.8 Preview is the model behind Kimi Code's default kimi-for-coding ID. It has a 1M-token context on every paid plan, reads images and video, and is described by Moonshot as close to K3 with more efficient thinking. Here's what's confirmed, what isn't, and when to use it.

kimi-for-coding 1M context Plus plan and up
Kimi Code > /model k3.1-proK3.1 Prok3.1-pro-highspeedK3.1 Pro Highspeedk3K3k3-256kK3 (256K)kimi-for-codingK2.8 Previewkimi-for-coding-highspeedK2.7 Code Highspeed K2.8 PreviewContext1MThinkingmaxInputtext, image, videoPlanPlus and up“Close to K3” (Moonshot) >Fix the failing checkout test ● thinking (max) … Read src/cart/discount.ts · tests/checkout.test.ts Edit discount.ts - total + off+ total - off Run npm test ✓ 12 tests passed Example session for illustration
OVERVIEW

K2.8 Preview at a glance

11 Sep2026 release

Rolled out to all Kimi Code users under the existing ID.

1MToken context

On every eligible plan, including Plus.

VisionImage and video

Paste screenshots, mock-ups or screen recordings.

≈ K3Moonshot's claim

“Close to K3” with more efficient thinking than K2.7 Code.

In short: if you use Kimi Code with the default settings, you're already using K2.8 Preview. It replaced K2.7 Code behind kimi-for-coding, so no settings change was needed.
SPECS

Kimi K2.8 Preview specifications

SpecK2.8 Preview
Released11 September 2026, in Kimi Code
Model IDkimi-for-coding (the same ID K2.7 Code used before)
StatusPreview: no stable release date announced
Context windowUp to 1,048,576 tokens (1M) on every eligible plan
InputText, images and video
Thinkinglow, high or max (default max); can be turned off
SpeedRegular speed
PlansKimi Plus and above (legacy: Andante and above)
WhereKimi Code CLI, Desktop, VS Code and other tools using a Kimi Code key
Kimi APINot listed on the pay-as-you-go Kimi API
Open weightsNone published
BenchmarksNone published; Moonshot says performance is close to K3

From Kimi Code's model configuration page and What's New notes. Moonshot hasn't published parameter counts or a technical report for K2.8.

LINEUP

Where K2.8 fits in Kimi Code

Kimi Code offers six model IDs. K2.8 Preview is the regular-speed everyday option and the only one with a full 1M context on the Plus plan.

Model IDModelContextPlansSpeed / quota
k3.1-proK3.1 Pro1M (Pro plan and up)Plus and upRegular
k3.1-pro-highspeedK3.1 Pro Highspeed1M (Pro plan and up)Pro and up2–3× speed, 2× quota
k3K31M (Pro plan and up)Plus and up1M uses 2× the quota of k3-256k
k3-256kK3 (256K)256KPlus and upRegular; images only
kimi-for-codingK2.8 Preview1M on all eligible plansPlus and upRegular
kimi-for-coding-highspeedK2.7 Code Highspeed256KPro and upAbout 6× speed, 3× quota

From Kimi Code's model table, November 2026. More on each: K3.1 Pro, K3.1 Pro Highspeed, K3 and K2.7 Code.

CHOOSE

K2.8 Preview vs K3 and K3.1 Pro

  • 1
    Pick K2.8 Preview for everyday coding. Kimi Code recommends it for code completion and routine development. Its thinking is more efficient, so jobs tend to finish sooner.
  • 2
    Pick K3.1 Pro for the hardest work. Big refactors, new features across many files and tricky bugs. It's Kimi's strongest model, with published benchmark results.
  • 3
    Pick a Highspeed model when you're waiting. K3.1 Pro Highspeed or K2.7 Code Highspeed reply faster but use 2–3× the quota.
  • 4
    On the Plus plan with a huge repo, K2.8 is the model that gives you the full 1M context.
WHICH KIMI CODE MODEL? Bug fixes, tests, completionsEveryday work K2.8 Previewkimi-for-coding Big refactors, new featuresHard, multi-file work K3.1 Prok3.1-pro Quick edits, live pairingYou're waiting on it Highspeedk3.1-pro-highspeed Huge repo on the Plus planNeeds 1M context K2.8 Previewkimi-for-coding Our suggestions, based on Kimi Code’s model notes.

A simple way to split work between Kimi Code's models.

TESTS

Early test results

There are no formal benchmarks yet, only a few hands-on reports. Treat them as early signals, not proof.

DateSourceWhat they found
11 Sep 2026Developer on X (@notjazii)Same task as K3 at max thinking: K2.8 took 80 minutes and about $10 but needed more than one attempt; K3 got it in one go in 120 minutes for about $14.
12 Sep 2026kimi-ai.freeFixed a discount-calculation bug without changing the function signature, and the existing unit test passed.
Sep 2026Several reviewsAgree it's a sensible default for everyday coding, with K3 or K3.1 Pro kept for the hardest tasks.

Single tasks from individual testers. Your results will depend on your codebase and prompts.

SETUP

How to use K2.8 Preview

In Kimi Code

Install Kimi Code, sign in with a Plus plan or higher, and type /model. Choose kimi-for-coding, which is the default.

In other tools

Point an OpenAI- or Anthropic-compatible coding tool at Kimi Code's endpoint with your Kimi Code key and the model kimi-for-coding.

Set the effort

Max is the default. Use high or low for simple tasks to get faster replies and save quota.

# Kimi Code CLI (Plus plan and up) /model # choose kimi-for-coding → K2.8 Preview # Thinking effort: low | high | max (default max) # Lower effort = faster replies and less quota for simple jobs

Full install guide: Kimi Code. Plan prices: Kimi pricing. In China, use api.kimi.com instead of api.kimi.ai.

LIMITS

Limitations to know

It's a preview

Moonshot can update the model behind kimi-for-coding without changing the ID, so behaviour can shift. Re-test before relying on results.

No published benchmarks

Moonshot has only said performance is “close to K3”. There are no official scores and no independent leaderboard results yet.

Kimi Code only

There's no official pay-as-you-go API price or open weights. Third-party gateways that list it are unofficial.

Watch the fallbacks

With thinking turned off, Kimi Code serves K2.8 without thinking, even if you picked K3. A mistyped model ID also falls back to it quietly.

FAQ

Kimi K2.8 Preview FAQ

Kimi K2.8 Preview is a coding and agent model from Moonshot AI, released in Kimi Code on 11 September 2026. It runs behind the kimi-for-coding model ID, has a 1M-token context and accepts images and video. Moonshot says its overall performance is close to Kimi K3, with more efficient thinking than K2.7 Code.

It uses kimi-for-coding, the same ID that K2.7 Code used before. Tools already set to kimi-for-coding switched to K2.8 Preview automatically on 11 September 2026.

Moonshot describes K2.8 Preview as close to K3, not better. In one developer's test, K2.8 finished faster and cheaper but needed more attempts than K3. Use K2.8 for everyday coding and K3.1 Pro for the hardest work.

K2.8 Preview is available in Kimi Code on the Plus plan and above (Andante and above on legacy plans). The free plan doesn't include Kimi Code.

Yes. Kimi Code lists K2.8 Preview with up to 1M tokens on every eligible plan, while K3 and K3.1 Pro only get the full 1M on the Pro plan and above.

Not on Moonshot's pay-as-you-go Kimi API, which lists K3.1 Pro and K3. You can use it in Kimi Code and in other coding tools that connect with a Kimi Code key.

No. Moonshot hasn't published weights or a licence for K2.8 Preview. K2.7 Code, the model it replaced, did have open weights.

Moonshot hasn't announced a date for a stable K2.8 release. We'll update this page and our Kimi updates timeline when it does.

Code with K2.8 Preview

It's the default model in Kimi Code on the Plus plan and above.