OpenAI and Anthropic-compatible APIs

Inference infrastructure, without the guesswork.

EU-hosted, OpenAI-compatible inference for chat, reasoning, vision, speech recognition, and speech synthesis.

Start building View models API reference

Make your first request¶

Set KOSCOMPUTE_API_KEY, then call the standard OpenAI-compatible base URL.

curl https://api.koscompute.com/v1/chat/completions \
  -H "Authorization: Bearer $KOSCOMPUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3.8-27b",
    "messages": [{"role": "user", "content": "Explain edge inference in one sentence."}],
    "max_completion_tokens": 120
  }'

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["KOSCOMPUTE_API_KEY"],
    base_url="https://api.koscompute.com/v1",
)
response = client.chat.completions.create(
    model="qwen/qwen3.8-27b",
    messages=[{"role": "user", "content": "Explain edge inference in one sentence."}],
    max_completion_tokens=120,
)
print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.KOSCOMPUTE_API_KEY,
  baseURL: "https://api.koscompute.com/v1",
});
const response = await client.chat.completions.create({
  model: "qwen/qwen3.8-27b",
  messages: [{ role: "user", content: "Explain edge inference in one sentence." }],
  max_completion_tokens: 120,
});
console.log(response.choices[0].message.content);

Live model catalog
Fetched once on load; build snapshot available as fallback.

One endpoint, several workflows¶

Chat and codeStreaming, tools, structured JSON, and usage. VisionHTTPS images and data URLs with model-specific limits. AudioSpeech transcription and raw synthesized audio.

Privacy by design¶

KosCompute returns X-Zero-Data-Retention: true for API traffic. Request bodies, prompts, and generated content are not retained by the router. Operational and billing metadata can include request IDs, model, timing, status, and token counts. Read the Zero Data Retention contract for the precise boundary.