Kimi K3
Where it stands
Quality in independent tests against price, next to 135 other models. Up means more capable, left means cheaper.
About this model
Kimi K3 is an AI model from Moonshot AI, made available through the API on July 16, 2026. It can read up to 1,048,576 tokens in one request, roughly 1,966 pages of text, and write up to 943,718 tokens in a single answer. Besides text, it accepts images and video as input.
Through the API it costs $3 per million input tokens and $15 per million output tokens. It is among the more expensive models we track, a price usually reserved for the most capable, flagship-class models.
It supports extended reasoning: it can "think" step by step before answering, which helps on maths, code and complex questions. It can call tools and functions, so developers can connect it to search, databases or other software. Its weights are openly published (moonshotai/Kimi-K3), so it can also be downloaded and run on your own hardware.
Quality scores
Rankings →- GPQA DiamondVery hard multiple-choice questions in biology, physics and chemistry, written by PhD experts. Random guessing scores 25%.93.1%#14 of 116
- AIME mathCompetition-level math problems in the style of the American Invitational Mathematics Examination. Share solved.97.2%#20 of 112
- FrontierMathOriginal, unpublished math problems written by professional mathematicians, from advanced to research level. Share solved.72.2%#18 of 73
- SimpleQA VerifiedShort factual questions. Measures how often the model gets facts right instead of making something up.50.6%#21 of 69
Where to use it
Companies that serve this model through an API: 22. Prices per million tokens, cheapest first.
| Provider | Inputi | Outputi | Contexti | Uptime (24 h)i |
|---|---|---|---|---|
InferenceNetCheapestfp4 | $0.40 | $9 | 1M | 99.5% |
Relacefp4 | $0.40 | $10 | 1M | 97.6% |
Wafer | $1.44 | $9 | 1M | 99.8% |
Morphfp8 | $1.78 | $11.29 | 1M | 99.2% |
Sail Researchfp4 | $0.61 | $12.95 | 1M | 99.8% |
Makora | $2.04 | $12.75 | 1M | 98.5% |
Phala | $2.55 | $12.75 | 1M | 99.3% |
DigitalOcean | $2.55 | $12.95 | 1M | 88.9% |
Wafer | $2.80 | $14 | 1M | 99.7% |
DeepInfrabf16 | $2.85 | $14.25 | 1M | 98.9% |
Decartmxfp4 | $3 | $15 | 1M | 75.7% |
AkashMLfp4 | $3 | $15 | 1M | 99.3% |
Chutesmxfp4 | $3 | $15 | 1M | 96.7% |
Parasailfp4 | $3 | $15 | 1M | 99.0% |
Modalmxfp4 | $3 | $15 | 1M | 99.5% |
Together | $3 | $15 | 1M | 98.5% |
Fireworks | $3 | $15 | 1M | 99.8% |
BaseTenfp8 | $3 | $15 | 1M | 97.7% |
Moonshot AIOfficialmxfp4 | $3 | $15 | 1M | 98.8% |
Fireworks | $3.30 | $16.50 | 1M | 99.5% |
Alibaba | $3.45 | $17.25 | 1M | 96.3% |
Fireworks | $4.50 | $22.50 | 1M | 99.5% |
Availability and prices come from OpenRouter’s public API and are refreshed daily. Uptime is the share of successful requests over the last 24 hours. Some providers run a compressed version of open models (fp8, int4), which is cheaper but can be slightly less accurate.
Use it in your code
Call Kimi K3 through OpenRouter’s OpenAI-compatible API: create a key on openrouter.ai, then run one of these examples.
curl https://openrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $OPENROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "moonshotai/kimi-k3", "messages": [{"role": "user", "content": "Hello! What can you do?"}]}'import os
from openai import OpenAI
client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=os.environ["OPENROUTER_API_KEY"])
reply = client.chat.completions.create(
model="moonshotai/kimi-k3",
messages=[{"role": "user", "content": "Hello! What can you do?"}],
)
print(reply.choices[0].message.content)import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://openrouter.ai/api/v1", apiKey: process.env.OPENROUTER_API_KEY });
const reply = await client.chat.completions.create({
model: "moonshotai/kimi-k3",
messages: [{ role: "user", content: "Hello! What can you do?" }],
});
console.log(reply.choices[0].message.content);The same code works with every model on this site: only the model ID changes. Many labs also offer their own API and SDK.
What it costs in practice
- 1,000 chat replies$6.00
- Summarise a 100-page document$0.165
- An app with 10,000 requests a day for a month$2,700
Similar-priced alternatives
| Model | Inputi | Outputi | Contexti | |
|---|---|---|---|---|
Ember-1Fireworks AI | $3 | $15 | 1M | Compare with ⇄ |
Aion 3.5Aion Labs | $3 | $6 | 256K | Compare with ⇄ |
GLM 5.3 PrimeZ.ai (Zhipu) | $2.80 | $8.80 | 1M | Compare with ⇄ |
Qwen3.8 Max PrimeAlibaba Qwen | $4 | $12 | 1M | Compare with ⇄ |
Claude Opus 5.5Anthropic | $4 | $20 | 1M | Compare with ⇄ |
Sonar Pro SearchPerplexity | $3 | $15 | 200K | Compare with ⇄ |
News about Moonshot AI
See all →Frequently asked questions
- How much does Kimi K3 cost?
- Kimi K3 costs $3 per million input tokens and $15 per million output tokens through the API. For example, 1,000 chat replies of about 500 tokens in and 300 out cost around $6.00.
- What is the context window of Kimi K3?
- Kimi K3 can read up to 1,048,576 tokens in one request, about 1,966 pages of text. It can write up to 943,718 tokens in a single answer.
- Is Kimi K3 open source?
- Yes. Its weights are published (moonshotai/Kimi-K3), so you can download it and run it on your own hardware, or use it through one of 22 API providers.
- Who makes Kimi K3?
- Kimi K3 is made by Moonshot AI. It became available through the API on July 16, 2026.
- How good is Kimi K3?
- Kimi K3 has an overall index of 157.7 in Epoch AI’s independent tests, ranking #12 of 136 models we track.
- Can Kimi K3 read images?
- Yes. Besides text, it accepts images and video as input.
- What are the alternatives to Kimi K3?
- Models with a similar price from other companies include Ember-1, Aion 3.5 and GLM 5.3 Prime. You can compare them side by side on this page.