Models·MoonshotAI

Kimi K2

moonshotai/kimi-k2

Kimi K2 is MoonshotAI’s open-weight text model with tool support and a 131,072-token context window.

$0.57Input, per million tokens
$2.30Output, per million tokens
131KContext window
98KLongest answer

131K tokens of context — about 98,304 words, on a scale that ends at 2M.

What it can do

  • Tool use
  • Open weights
  • Takes text
  • Gives text

Released 11 Jul 2025 · checked 3 Oct 2026.

About it

Kimi K2 is MoonshotAI’s Kimi K2 Instruct model, released July 11, 2025. It uses a mixture-of-experts design with 1 trillion total parameters and 32 billion active for each processing step. The model handles text, supports tools, has open weights, and accepts up to 131,072 tokens at once.

Review

In a line

Kimi K2 is MoonshotAI’s large text model, also listed as Kimi K2 Instruct. It has open weights and supports tool use.

The model is identified as moonshotai/kimi-k2, with the Hugging Face name moonshotai/Kimi-K2-Instruct.

What it's good at

This model suits text-based work where tool calling matters. Tool calling lets software ask the model to trigger supported actions rather than only return text.

Its mixture-of-experts design has 1 trillion total parameters, with 32 billion active during each processing step. It works with text input and produces text output. See other models with tool calling.

Price in plain terms

Input costs $0.57 per million tokens, and output costs $2.3 per million tokens. Tokens are small pieces of text counted for billing.

Reading a page of about 500 words costs about 0.037 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.17 cents. One thousand such turns cost about $1.72.

Prices are listed per million tokens, so your total depends on how much text the model reads and writes.

Context and limits

The context window is 131,072 tokens, so one request can include a large amount of text and conversation history. The maximum output is 98,304 tokens in one answer.

Kimi K2 is text-only: its listed inputs and outputs are text.

Ways to use it

Use the model with tool support or in an open-weight setup. Its model ID is moonshotai/kimi-k2, and its Hugging Face ID is moonshotai/Kimi-K2-Instruct.

MoonshotAI places it in the moonshotai model family. You can compare it with other MoonshotAI models and open-weight models.

Verdict

Choose Kimi K2 if you need a text model with a very large context window, open weights, and tool support. Its input and output token rates differ, with rough costs supplied for common chat use.

Skip it if your work requires image or audio input, since the listed modality is text only.

Verdict

Kimi K2 is suited to text workflows that can use tool calling and long prompts. Open weights provide an option for open-weight deployment. The model’s text-only design rules out image and audio tasks, and its cost depends on usage. Compare the input and output rates with your expected traffic, especially if the model will generate much more text than it reads.

  • Open weights are available.
  • Supports tool calling.
  • 131,072-token context window.
  • 98,304-token maximum output.
  • Rough cost estimates are provided for typical chat turns.
  • Text input and output only.
  • Output tokens cost more than input tokens.
  • Pick it ifText workflows needing tool calling; Long documents and conversation context; Buyers comparing per-token model prices
  • Skip it ifYou need image or audio handling.; Your software cannot work with tool-calling models.

Pricing

Billing is per million tokens: $0.57 for input and $2.3 for output. Roughly, 1,000 chat turns of the stated size cost about $1.72.

Questions

Who makes Kimi K2?

MoonshotAI makes it. The model belongs to the moonshotai family.

Is Kimi K2 open source?

Its weights are listed as open weights. The Hugging Face ID is moonshotai/Kimi-K2-Instruct.

Can it use tools?

Yes. Tool support is listed for this model.

How much text can it read at once?

Its context window is 131,072 tokens.

How much can it write in one answer?

The maximum output is 98,304 tokens.

What does it cost?

Input is $0.57 per million tokens and output is $2.3 per million tokens. A typical 1,000-turn estimate in the facts is about $1.72.

Does it handle images?

No image or audio capability is listed. Its modality and inputs are text, and its output is text.

On these boards

More from MoonshotAI