Kimi K3
moonshotai/kimi-k3
A multimodal reasoning model for complex coding, knowledge work and long-horizon agent workflows. Combine text, images and video with a million-token context window. Good for complex coding, knowledge work and agent workflows.
Our discountYour discount
−10%off the official price−15%off the official priceUp to −20%off the official priceTier 3
- Tier 1−10%
- Tier 2−15%
- Tier 3−20%
Official $3.00 / $15.00Our price $2.70 / $13.50$2.55 / $12.75$2.40 / $12.00 Per 1M tokensTier 3: up to −20%
- Inputtext + image + videoOutput: text
- Context1.05M tokensUp to 944K output per reply
- ReleasedJuly 16, 2026In the OceanToken catalog
01 · Learn
What is Kimi K3?
Kimi K3 is a multimodal language model from Moonshot AI on the OceanToken API, with a 1.05M-token context window and reasoning, tool calling, structured outputs and image input.
- −20%Our price vs official$2.40 / $12.00 Per 1M tokens · Official: $3.00 / $15.00
- 1.05MContext windowUp to 944K output per reply
- text + image + videoInputOutput: text
- Up to 60% offEnterprise rate vs officialUp to, by quote
What Kimi K3 can do.
- Complex coding
- Knowledge work
- Agent workflows
02 · Compare
Is Kimi K3 cheaper than similar models?
The official price, then our price, for Kimi K3 and the models most like it.
OfficialOur price
Kimi K3MoonshotAI
$9.00$7.20−20%up to −60%Gemini 3.1 Pro PreviewGoogle
$7.00$6.30−10%up to −30%Claude Opus 5.5Anthropic
$12.00$10.80−10%up to −30%GPT-6.1 SolOpenAI
$6.00$5.40−10%up to −30%Qwen3.8 MaxQwen
$4.00$3.60−10%up to −60%GLM 5.3Z.ai
$2.90$1.74−40%up to −60%
Per 1M tokens, half input and half output. Prices as of November 18, 2026. Enterprise rate: quote by email, we activate it for you.
Kimi K3 benchmarks.
- 1,488LMArena TextHuman preference rating, LMArena
Scores from independent leaderboards, as of October 27, 2026. We did not run these tests ourselves.
Kimi K3 specs.
- Context window
- 1.05M tokens (1,048,576)
- Max output
- 944K tokens (943,718)
- Input
- text, image, video
- Output
- text
- Reasoning
- Yes,
reasoning_effort - Tool calling
- Yes,
tool_choice - Structured outputs
- Yes,
response_format - Prompt caching
- Yes, at the official price
- Released
- July 16, 2026
Model ID and endpoint
- OceanToken model ID
moonshotai/kimi-k3- OpenAI-compatible base URL
https://api.oceantoken.ai/v1- Endpoint
POST /v1/chat/completions
03 · Buy and build
How much does the Kimi K3 API cost?
Kimi K3 costs $2.70 per 1M input tokens and $13.50 per 1M output tokens. Cached input costs $0.30 per 1M tokens. The official price is $3.00 per 1M input tokens and $15.00 per 1M output tokens. At the top cumulative top-up tier it costs $2.40 per 1M input tokens and $12.00 per 1M output tokens. Blended (50% input / 50% output): official $9.00 → top tier $7.20 (−20%). These are OceanToken starting prices as of November 18, 2026, charged per use to your prepaid balance. The console always shows the current price.
| Rate | Input | Output | Off |
|---|---|---|---|
| Official | |||
| Tier 1From sign-up | $2.70 | $13.50 | −10% |
| Tier 2$1,000 total | $2.55 | $12.75 | −15% |
| Tier 3$10,000 total | $2.40 | $12.00 | −20% |
| Enterprise rateBy quote | up to −60% | ||
Thinking tokens are billed at the output price.
Per 1M tokens. Half input, half output: official $9.00, ours $7.20. Prompt cache and web search are billed at the official price, at every tier.
How tiers work
- Self-serveTier 1From sign-up
- Self-serveTier 2$1,000 total
- Self-serveTier 3$10,000 total
- By quoteEnterprise rateUp to 60% off, by quote
You pay per use from a prepaid balance, with no subscription. Official price: the model maker’s list price. Tier 1 to 3 are standing rates by total top-up.
How to use Kimi K3 on OceanToken.
Three steps from no account to a working request. The same key and balance work for every model in the catalog.
Create an account and get an API key
One key works for every model your account can use.
Get API KeyAdd credits
OceanToken is prepaid: add credits, and each request is billed from that balance.
Add creditsCall Kimi K3 from your code
Point an OpenAI SDK at
See the API request in docshttps://api.oceantoken.ai/v1and use the model IDmoonshotai/kimi-k3.
client.pyPythonfrom openai import OpenAI
# Your API Key from the OceanToken console
client = OpenAI(base_url="https://api.oceantoken.ai/v1",
api_key="YOUR_API_KEY")
reply = client.chat.completions.create(
model="moonshotai/kimi-k3",
messages=[{"role": "user", "content": "Hello"}])
print(reply.choices[0].message.content)client.tsTypeScriptimport OpenAI from "openai";
// Your API Key from the OceanToken console
const client = new OpenAI({ baseURL: "https://api.oceantoken.ai/v1", apiKey: "YOUR_API_KEY" });
const reply = await client.chat.completions.create({
model: "moonshotai/kimi-k3",
messages: [{ role: "user", content: "Hello" }],
});
console.log(reply.choices[0].message.content);request.mjsJavaScript// Your API Key from the OceanToken console
const KEY = "YOUR_API_KEY";
const res = await fetch("https://api.oceantoken.ai/v1/chat/completions", {
method: "POST",
headers: { Authorization: `Bearer ${KEY}`, "Content-Type": "application/json" },
body: JSON.stringify({ model: "moonshotai/kimi-k3", messages: [{ role: "user", content: "Hello" }] }),
});
const reply = await res.json();
console.log(reply.choices[0].message.content);request.shcurl# Your API Key from the OceanToken console
curl https://api.oceantoken.ai/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "moonshotai/kimi-k3",
"messages": [{"role": "user", "content": "Hello"}]}'main.goGopackage main
import (
"fmt"
"io"
"net/http"
"strings"
)
func main() {
// Your API Key from the OceanToken console
body := strings.NewReader(`{"model": "moonshotai/kimi-k3", "messages": [{"role": "user", "content": "Hello"}]}`)
req, _ := http.NewRequest("POST", "https://api.oceantoken.ai/v1/chat/completions", body)
req.Header.Set("Authorization", "Bearer YOUR_API_KEY")
req.Header.Set("Content-Type", "application/json")
res, err := http.DefaultClient.Do(req)
if err != nil {
panic(err)
}
defer res.Body.Close()
out, _ := io.ReadAll(res.Body)
fmt.Println(string(out))
}Main.javaJavaimport java.net.URI;
import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
public class Main {
public static void main(String[] args) throws Exception {
// Your API Key from the OceanToken console
String body = "{\"model\": \"moonshotai/kimi-k3\", \"messages\": [{\"role\": \"user\", \"content\": \"Hello\"}]}";
HttpRequest req = HttpRequest.newBuilder(URI.create("https://api.oceantoken.ai/v1/chat/completions"))
.header("Authorization", "Bearer YOUR_API_KEY")
.header("Content-Type", "application/json")
.POST(HttpRequest.BodyPublishers.ofString(body))
.build();
HttpResponse<String> res = HttpClient.newHttpClient().send(req, HttpResponse.BodyHandlers.ofString());
System.out.println(res.body());
}
}request.phpPHP<?php
// Your API Key from the OceanToken console
$ch = curl_init("https://api.oceantoken.ai/v1/chat/completions");
curl_setopt_array($ch, [
CURLOPT_POST => true,
CURLOPT_RETURNTRANSFER => true,
CURLOPT_HTTPHEADER => ["Authorization: Bearer YOUR_API_KEY", "Content-Type: application/json"],
CURLOPT_POSTFIELDS => '{"model": "moonshotai/kimi-k3", "messages": [{"role": "user", "content": "Hello"}]}',
]);
echo curl_exec($ch);request.rbRubyrequire "net/http"
# Your API Key from the OceanToken console
uri = URI("https://api.oceantoken.ai/v1/chat/completions")
req = Net::HTTP::Post.new(uri, "Authorization" => "Bearer YOUR_API_KEY", "Content-Type" => "application/json")
req.body = '{"model": "moonshotai/kimi-k3", "messages": [{"role": "user", "content": "Hello"}]}'
res = Net::HTTP.start(uri.host, uri.port, use_ssl: true) { |http| http.request(req) }
puts res.bodyKimi K3 API: frequently asked questions.
What is Kimi K3?
A multimodal reasoning model for complex coding, knowledge work and long-horizon agent workflows. Combine text, images and video with a million-token context window.
How much does Kimi K3 cost?
Kimi K3 costs $2.70 per 1M input tokens and $13.50 per 1M output tokens. Cached input costs $0.30 per 1M tokens. The official price is $3.00 per 1M input tokens and $15.00 per 1M output tokens. At the top cumulative top-up tier it costs $2.40 per 1M input tokens and $12.00 per 1M output tokens. Blended (50% input / 50% output): official $9.00 → top tier $7.20 (−20%). These are OceanToken starting prices as of November 18, 2026, charged per use to your prepaid balance. The console always shows the current price.
What is the context length of Kimi K3?
Kimi K3 has a context window of 1.05M tokens (1,048,576 tokens).
Does Kimi K3 support tool calling, vision and reasoning?
Kimi K3 supports tool calling, structured outputs, image input (vision) and reasoning. It takes text, image and video input and returns text.
How do I call Kimi K3 through the API?
Send requests to https://api.oceantoken.ai/v1/chat/completions with the model ID moonshotai/kimi-k3 and your OceanToken API key. The API is OpenAI-compatible, so an OpenAI SDK needs only the base URL https://api.oceantoken.ai/v1 and your key. The Responses API and the Anthropic Messages API (/v1/messages) work too. The same OceanToken API key works for every model your account can use. Full reference: docs.oceantoken.ai.
How do I top up?
OceanToken is prepaid. Open Credits in the console, choose an amount and pay by card on Stripe's hosted checkout page (Link works too). Your balance updates a few seconds after payment. Have a trial code? Redeem it on the same page.
Do credits expire, and can I get a refund?
Purchased credits are valid for 12 months from the date of purchase. Credits from trial codes, promotions or gifts may have a shorter validity. All purchases are final and non-refundable, except where the law requires a refund. Section 6 of the Terms of Service has the details.
Am I charged for failed requests?
No. If a request fails, you are not charged: that covers a request refused for an insufficient balance or by a safety check, one whose provider fails before returning any output, and a video job that fails or is refused by a safety check, whose held amount goes back to your balance. The one exception: if a request breaks off after the model has started producing output, for example a stream cut off part-way, you pay for the tokens that were processed and generated for it.
Related articles
Similar language models
- Qwen3.8 FlashOur discount−20%
- Grok 4.7Our discount−30%
- Grok 4.6Our discount−30%
- Qwen4 MaxOur discount−20%
- DeepSeek V4 ProOur discount−40%
- GLM 5.3Our discount−40%
Compare Kimi K3 with other models.
Keep exploring
Sources
- OceanToken model catalog: every price, context window, input, output and parameter on this page (November 18, 2026).
- How tiers work: Tier 1 to 3 and the Enterprise rate.
- LMArena: benchmark scores (October 27, 2026).