Network Routes and Endpoints

How to choose the Inference Space mainland China and international routes, console, and SDK Base URLs.

Inference Space provides mainland China and international public routes. The protocol, model IDs, and authentication method are identical; choose the route closest to your users and servers.

Route Overview

Route	Console	OpenAI-Compatible Base	Anthropic-Compatible Base	Image / Audio Base
Mainland China route	`https://ai.tos.run`	`https://api.tos.run/v1`	`https://api.tos.run`	`https://api.tos.run`
International route	`https://ai.inf.space`	`https://ai.inf.space/v1`	`https://ai.inf.space`	`https://ai.inf.space`

The mainland China console uses ai.tos.run, while its API requests use api.tos.run. The international console and API use ai.inf.space.

If most of your users or servers are outside mainland China, prefer https://ai.inf.space. If your business is primarily in mainland China, prefer https://api.tos.run for API calls.

Common SDK Configuration

Mainland China route:

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://ai.inf.space/v1",
  apiKey: process.env.TOS_API_KEY,
});

International route:

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://ai.inf.space/v1",
  apiKey: process.env.TOS_API_KEY,
});

Mainland China route:

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://ai.inf.space",
  authToken: process.env.TOS_API_KEY,
});

International route:

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://ai.inf.space",
  authToken: process.env.TOS_API_KEY,
});

Mainland China route:

curl "https://ai.inf.space/v1/chat/completions" \
  -H "Authorization: Bearer $TOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen-plus","messages":[{"role":"user","content":"Hello"}]}'

International route:

curl "https://ai.inf.space/v1/chat/completions" \
  -H "Authorization: Bearer $TOS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen-plus","messages":[{"role":"user","content":"Hello"}]}'

Endpoint Selection

Capability	Mainland China route	International route
OpenAI Chat Completions	`POST https://ai.inf.space/v1/chat/completions`	`POST https://ai.inf.space/v1/chat/completions`
OpenAI Responses	`POST https://ai.inf.space/v1/responses`	`POST https://ai.inf.space/v1/responses`
Anthropic Messages	`POST https://ai.inf.space/v1/messages`	`POST https://ai.inf.space/v1/messages`
Image generation	`POST https://ai.inf.space/v1/images/generations`	`POST https://ai.inf.space/v1/images/generations`
Image editing	`POST https://ai.inf.space/v1/images/edits`	`POST https://ai.inf.space/v1/images/edits`
Speech recognition	`POST https://ai.inf.space/v1/audio/transcriptions`	`POST https://ai.inf.space/v1/audio/transcriptions`
Speech synthesis	`POST https://ai.inf.space/v1/audio/speech`	`POST https://ai.inf.space/v1/audio/speech`

Notes

All API Keys use Authorization: Bearer gk_...; see Authentication and API Keys.
Use https://ai.inf.space/v1 for OpenAI-compatible APIs and https://ai.inf.space for Anthropic-compatible APIs.
The English public entry point does not use a separate api.* domain.
If the same Key returns 401 on one route, confirm that the organization and API Key have been enabled for that route.

Route Overview

Common SDK Configuration

Endpoint Selection

Notes

On this page