Submit
Moonshot AIOpen-weight·1049K context·— params·Kimi K3 License

Kimi K3moonshot/kimi-k3

Kimi K3 is Moonshot AI's frontier model, released 2026-07-16. At 2.8 trillion total parameters (104B activated) it is the largest open-weight model published to date, with a 1,048,576-token context window, native visual understanding, and an always-on reasoning mode. Weights ship under the custom Kimi K3 License with MXFP4 quantization-aware training. Pricing is the highest of any Chinese API in this directory: cache misses cost $3.00 per 1M input tokens, but cache hits drop to $0.30, so prompt-caching discipline changes the economics by 10x.

Cheapest blended:$6.00 / 1M tokenson Moonshot AI · 1 provider listed

Pricing across providers

Sort by:
ProviderInput /1MOutput /1MBlended /1MLatency p50FormatFreshnessAction
Moonshot AI
kimi-k3
$3.00$15$6.00—OpenAI-compatiblePrice 1mo agoTry →

Affiliate disclosure: We may earn a commission from qualified signups. Pricing independence is enforced at the data layer — see our Editorial Independence Policy.

Works with

Point any of these clients at a hosting's base URL — they all speak at least one of this model's endpoint protocols (OPENAI_COMPATIBLE).

Capabilities

  • chat
  • code
  • reasoning
  • long_context
  • agent
  • vision

Code samples

Example using Moonshot AI — the cheapest hosting for this model as of last verification. Swap base_url and model to use a different provider from the matrix above.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://api.moonshot.cn/v1",
)

response = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

Technical specs

Context
1049K
Max output
—
Parameters
—
Release
—
Training cutoff
—
License
Kimi K3 License
Hugging Face →

Similar models

Compare with

  • Kimi K3 vs DeepSeek V3.1
    Comparison planned — not yet published
  • Kimi K3 vs R1
    Comparison planned — not yet published
  • Kimi K3 vs DeepSeek V3
    Comparison planned — not yet published

Frequently asked

How much does Kimi K3 cost?+
The cheapest public hosting is $6.00 per 1M blended tokens on Moonshot AI. 1 total providers are listed above with per-input / per-output / cached pricing.
How do I access Kimi K3 from outside China?+
All hostings listed above support global access. The official API (e.g. api.deepseek.com, dashscope-intl.aliyuncs.com) accepts international credit cards and does not require a Chinese mobile number. For privacy-sensitive workloads, third-party aggregators like Together.ai host the model on US/EU infrastructure.
Is Kimi K3 open-source? Can I fine-tune it?+
Yes. Kimi K3 is open-weight under the Kimi K3 License license. Weights are available on Hugging Face for local inference, fine-tuning, and commercial use (see license for specific terms).
Is Kimi K3 OpenAI-compatible?+
Most listed hostings expose an OpenAI-compatible API, so you can point an existing openai SDK client at the Provider's base_url and use the Provider's model name. See the Code Samples above for a copy-pasteable example.
What's the maximum context window for Kimi K3?+
The model supports up to 1,048,576 tokens of context (input + output). Some hosted versions may impose a smaller limit — check the "Context" column in the pricing matrix for each provider.