Anbindung in zwei Zeilen.

Schnellstart

curl https://cheapllm.shop/v1/chat/completions \
  -H "Authorization: Bearer $CHEAPLLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-1017",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Unter Windows in PowerShell:

Invoke-RestMethod -Method Post -Uri "https://cheapllm.shop/v1/chat/completions" `
  -Headers @{ Authorization = "Bearer $env:CHEAPLLM_API_KEY" } -ContentType "application/json" `
  -Body '{"model": "deepseek-v4.1-flash-1017", "messages": [{"role": "user", "content": "Hello!"}]}'

Python

from openai import OpenAI

client = OpenAI(base_url="https://cheapllm.shop/v1", api_key="sk-cheap-...")

resp = client.chat.completions.create(
    model="deepseek-v4.1-flash-1017",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

Node / TypeScript

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://cheapllm.shop/v1", apiKey: process.env.CHEAPLLM_API_KEY });

const resp = await client.chat.completions.create({
  model: "deepseek-v4.1-flash-1017",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(resp.choices[0].message.content);

Kein SDK, nur fetch

const r = await fetch("https://cheapllm.shop/v1/chat/completions", {
  method: "POST",
  headers: { "Authorization": `Bearer ${KEY}`, "Content-Type": "application/json" },
  body: JSON.stringify({ model: "deepseek-v4.1-flash-1017", messages: [{ role: "user", content: "Hello!" }] }),
});
const data = await r.json();

Keys

Sende deinen Schlüssel als Authorization: Bearer sk-cheap-.... Der Header x-api-key funktioniert ebenfalls, für Clients, die ihn verwenden.

• Erstelle bis zu 25 Schlüssel im Dashboard, einen pro App oder Teammitglied.
• Ein Schlüssel wird nur einmal angezeigt, wenn du ihn erstellst. Wir speichern nur einen Fingerabdruck.
• Das Widerrufen wirkt sich auf die nächste Anfrage aus.
• CORS ist offen, sodass Browser-Apps die API direkt aufrufen können. Bette einen Schlüssel aber nicht in eine öffentliche Webseite ein; jeder kann ihn lesen.

Chat-Completions

POST /v1/chat/completions nimmt denselben Body wie der OpenAI-Endpunkt. Die am häufigsten genutzten Felder:

Antworten enthalten ein usage-Objekt mit prompt_tokens und completion_tokens sowie einen x-request-id-Header, den es sich zu zitieren lohnt, wenn du uns schreibst.

Kontextfenster: Tokens.

Streaming

Mit "stream": true erhältst du text/event-stream: eine data: {...}-Zeile pro Chunk und ein abschließendes data: [DONE]. Füge "stream_options": {"include_usage": true} hinzu, um einen letzten Chunk mit Token-Zählungen zu erhalten.

stream = client.chat.completions.create(
    model="deepseek-v4.1-flash-1017",
    messages=[{"role": "user", "content": "Count to five, slowly."}],
    stream=True,
    stream_options={"include_usage": True},
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="", flush=True)
    elif chunk.usage:
        print("\n", chunk.usage)

Das Schließen der Verbindung stoppt die Generierung, sodass eine abgebrochene Anfrage keine Zeit mehr kostet.

Reasoning

Das Modell denkt nach, bevor es antwortet. Das Denken kommt getrennt von der Antwort an, sodass es deine Ausgabe nie verunreinigt:

Modelle

GET /v1/models listet auf, was auf der Karte steht. Heute ist das ein Modell:

{"object": "list", "data": [{"id": "deepseek-v4.1-flash-1017", "object": "model", "owned_by": "cheapllm"}]}

Model names containing deepseek are routed to it too, so tools with a hard-coded DeepSeek model name just work.

Limits

Free tier

During the beta every account is free, with no rate limit. When the beta ends, every account keeps a permanent free tier, with no card: at least 5,000 input tokens and 1,000 output tokens per hour (5,000 and 1,000 today), counted per account over a rolling hour, all keys together. These numbers can go up, never below that minimum. Past the free tier, usage is billed at $0.005 per million input tokens, $0.001 per million cached input tokens and $0.14 per million output tokens, with no rate limit. We'll email every account before the beta ends.

Rate and size limits

Errors

Errors come back as {"error": {"message", "type", "code"}}, like OpenAI's.

Anthropic and Responses formats

Tools built for the Anthropic Messages API (Claude Code, the Anthropic SDKs) can use the same model. Point them at https://cheapllm.shop; they call POST /v1/messages. The key works as x-api-key or Authorization: Bearer. Any model name is accepted and routed to deepseek-v4.1-flash-1017.

curl https://cheapllm.shop/v1/messages \
  -H "x-api-key: $CHEAPLLM_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model": "deepseek-v4.1-flash-1017", "max_tokens": 1024,
       "messages": [{"role": "user", "content": "Hello!"}]}'

Supported: text, images, system prompts, tool use and tool results, streaming events, stop sequences. Thinking blocks are not returned. POST /v1/messages/count_tokens returns an estimate.

Tools built for the OpenAI Responses API (Codex CLI, recent OpenAI SDK code) can call POST /v1/responses on https://cheapllm.shop/v1: text and image input, instructions, function and custom tools, streaming events. Stateless only (store and previous_response_id are not kept).

Your tools

Step-by-step settings for Claude Code, OpenCode, Codex CLI, DeepSeek Harness, Cline, Roo Code, Kilo Code, Continue, Copilot, Cursor, Zed and many more are on the integrations page. In short: anything with an "OpenAI-compatible" or "custom base URL" setting works with the base URL, your key and the model name above.

Environment variables

export OPENAI_BASE_URL=https://cheapllm.shop/v1
export OPENAI_API_BASE=https://cheapllm.shop/v1   # older libraries
export OPENAI_API_KEY=sk-cheap-...

Windows (PowerShell; setx keeps them for new windows):

setx OPENAI_BASE_URL "https://cheapllm.shop/v1"
setx OPENAI_API_BASE "https://cheapllm.shop/v1"
setx OPENAI_API_KEY "sk-cheap-..."

Or let the installer pick, install and configure your tools (all 40 from the integrations page): curl -fsSL https://cheapllm.shop/setup.sh | sh on macOS and Linux, irm https://cheapllm.shop/setup.ps1 | iex in Windows PowerShell.

Open WebUI

Admin settings, Connections, OpenAI API: add a connection with the base URL and your key. The model appears in the picker.

LibreChat

endpoints:
  custom:
    - name: "cheapllm"
      apiKey: "${CHEAPLLM_API_KEY}"
      baseURL: "https://cheapllm.shop/v1"
      models:
        default: ["deepseek-v4.1-flash-1017"]

Continue (VS Code, JetBrains)

models:
  - name: cheapllm
    provider: openai
    model: deepseek-v4.1-flash-1017
    apiBase: https://cheapllm.shop/v1
    apiKey: sk-cheap-...

Aider

aider --openai-api-base https://cheapllm.shop/v1 --openai-api-key sk-cheap-... \
      --model openai/deepseek-v4.1-flash-1017

Cursor

Settings, Models: paste your key in the OpenAI API key field, turn on "Override OpenAI Base URL" with the base URL, then add the model name as a custom model.

LangChain

from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="deepseek-v4.1-flash-1017", base_url="https://cheapllm.shop/v1", api_key="sk-cheap-...")

Vercel AI SDK

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
const cheapllm = createOpenAICompatible({ name: "cheapllm", baseURL: "https://cheapllm.shop/v1", apiKey: process.env.CHEAPLLM_API_KEY });
const model = cheapllm("deepseek-v4.1-flash-1017");

Stuck? Write to [email protected] with the x-request-id of the failing call.