Skip to content

Updated August 6, 2026

Every Kimi model, honestly compared.

Independent coverage of Moonshot AI’s Kimi family — verified specs, live API pricing and free tools that do the cost maths for you. No marketing copy, no invented benchmarks, every number sourced and dated.

Flagship
Kimi K3
Max context
1,048,576
Models tracked
7
kimi-api — python
# pip install openai
from openai import OpenAI

client = OpenAI(
    api_key="sk-...",
    base_url="https://api.moonshot.ai/v1",
)

response = client.chat.completions.create(
    model="kimi-k3",
    reasoning_effort="max",
    messages=[{
        "role": "user",
        "content": "Audit this repo.",
    }],
)

Models

Kimi models available today

Five Kimi models are open to new API users as of August 2026. Every figure below comes from Moonshot’s own platform documentation and is re-checked when a model ships.

  • Kimi K3

    Flagship

    Kimi K3 is Moonshot AI’s flagship model, with 2.8 trillion parameters, native visual understanding and a 1,048,576-token context window.

    Context
    1M
    Output / 1M
    $15.00
  • Kimi K2.7 Code is Moonshot AI’s coding-focused model, with a 262,144-token context window and higher instruction-following reliability in long contexts.

    Context
    256K
    Output / 1M
    $4.00
  • Kimi K2.7 Code HighSpeed is the same model as Kimi K2.7 Code served at roughly 180 tokens per second, at double the price.

    Context
    256K
    Output / 1M
    $8.00
  • Kimi K2.6 is a general-purpose model with a 262,144-token context window, supporting text, image and video input across thinking and non-thinking modes.

    Context
    256K
    Output / 1M
    $4.00
  • Kimi K4

    Unannounced

    Kimi K4 is an unannounced successor to Kimi K3. Moonshot AI has not confirmed its existence, specifications, pricing or release date.

    Tracking this page for the announcement.

Pricing

Kimi API pricing, side by side

USD per million tokens, excluding tax. Cache-hit pricing applies automatically to repeated prompt prefixes, which is why real-world bills land well below the headline rate.

Source: platform.kimi.ai model pricing pages, verified 6 August 2026.
ModelAPI IDContextInput (hit)Input (miss)Output
Kimi K3kimi-k31,048,576$0.30$3.00$15.00
Kimi K2.7 Codekimi-k2.7-code262,144$0.19$0.95$4.00
Kimi K2.7 Code HighSpeedkimi-k2.7-code-highspeed262,144$0.38$1.90$8.00
Kimi K2.6kimi-k2.6262,144$0.16$0.95$4.00
Kimi K2.5kimi-k2.5262,144$0.10$0.60$3.00

Changing now

Two Kimi models are being retired

Moonshot is consolidating the lineup around K3. If you are running either of these in production, you have a migration to plan.

Kimi K2.5

Closing

Kimi K2.5 is a multimodal model with a 262,144-token context window. It is closed to new API users and the platform sunsets it on August 31, 2026.

Sunsets August 31, 2026

Kimi K2

Discontinued

Kimi K2 was Moonshot AI’s open-weights model family from 2025. All K2 API variants were discontinued on May 25, 2026.

Shut down May 25, 2026

Tools

Free tools, no sign-up

Built because the answers were annoying to work out by hand. Everything runs in your browser — nothing you type is sent anywhere.

  • API cost calculator

    Enter your token volume and cache hit rate, get real monthly cost across every Kimi model.

  • Model comparison

    Filter and sort every Kimi model by context, price, speed and capability.

  • Context window visualizer

    See what 256K and 1M tokens actually hold — in pages, books and lines of code.

  • Token counter

    Paste any text and estimate its token count and cost before you send it.

FAQ

Common questions

Short answers to what people actually search for. Every figure links back to Moonshot’s own documentation.

What is the latest Kimi model?

Kimi K3 is the latest Kimi model, released by Moonshot AI on July 16, 2026. It has 2.8 trillion parameters, native visual understanding and a 1,048,576-token context window, and it is the flagship option on the Kimi API platform.

Is there a Kimi K4?

No. As of August 2026, Moonshot AI has not announced a model called Kimi K4. The current flagship is Kimi K3. Any specification, price or release date quoted for K4 elsewhere is speculation, not an announcement.

How much does the Kimi K3 API cost?

Kimi K3 costs $3.00 per million input tokens on a cache miss, $0.30 per million input tokens on a cache hit, and $15.00 per million output tokens. Prices exclude tax and are billed per token consumed.

Which Kimi model is cheapest?

Among models open to new API users, Kimi K2.6 has the lowest cache-hit input price at $0.16 per million tokens, with cache-miss input at $0.95 and output at $4.00 per million tokens. Kimi K2.5 is cheaper still but closes on August 31, 2026.

Is Kimi K2 still available?

No. Moonshot AI discontinued the entire kimi-k2 API series on May 25, 2026, including kimi-k2-thinking and the turbo variants. The open weights remain downloadable if you self-host, but the hosted API endpoints are gone.

What is the difference between Kimi K2.7 Code and K2.7 Code HighSpeed?

They are the same model. HighSpeed is served faster — roughly 180 tokens per second, up to 260 in short contexts — and costs exactly twice as much on every pricing tier. Choose it only when latency is your binding constraint.

Are Kimi models open source?

Moonshot AI publishes downloadable open weights for several Kimi models, including K3, K2.6 and K2. That lets you self-host, but it is not the same as an OSI-approved open-source licence — check the licence attached to each weight release before deploying commercially.