Skip to content
All models
MoonshotAvailable

Kimi K2

Kimi K2 Instruct is a large-scale Mixture-of-Experts (MoE) language model developed by Moonshot AI, featuring 1 trillion total parameters with 32 billion active per forward pass.

Try in ewpiremoonshotai/kimi-k2

Opens in the ewpire workspace. Creating an account is free; runs are paid from your credits.

Kimi K2 in plain words

Kimi K2 is a text model from Moonshot. It takes text as input and returns text. Its context window is 131,072 tokens. It costs 0.707471 credits per 1M input tokens and 2.854706 credits per 1M output tokens.

It is good at calling tools. It does not read images. It runs at medium speed among the models in the catalog. A context window of 131,072 tokens holds about 520,000 characters of English text, at 4 characters a token.

For example, a question of 2,000 tokens (about 8,000 characters) with an answer of 1,000 tokens (about 4,000 characters) costs 0.00427 credits. The exact charge follows the usage the model reports for your request.

Details

Made by
Moonshot
Good at
Tool use
Reads
Text
Produces
Text
Context window
131,072 tokens
Speed
Medium speed

Price

Input, per 1M tokens
0.707471 credits
Output, per 1M tokens
2.854706 credits

Charged per request from your credit balance. No subscription. See pricing

Use it through the API

Call it with the OpenAI SDK or plain HTTP: set the base URL to ewpire and pass this model id. Read the docs

curl https://ewpire.com/api/v1/chat/completions \
  -H "Authorization: Bearer $EWPIRE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"moonshotai/kimi-k2","messages":[{"role":"user","content":"Hello!"}]}'

The docs show how to list the models the API serves, with their capabilities and prices. Models in the docs

Similar models