Use OpenAI models with LaoZhang API

​

Models and endpoints

Chat Completions   POST https://api.laozhang.ai/v1/chat/completions
Responses          POST https://api.laozhang.ai/v1/responses
Image generation   POST https://api.laozhang.ai/v1/images/generations
Image editing      POST https://api.laozhang.ai/v1/images/edits
Transcription      POST https://api.laozhang.ai/v1/audio/transcriptions
Speech             POST https://api.laozhang.ai/v1/audio/speech
Embeddings         POST https://api.laozhang.ai/v1/embeddings

​

Configure the OpenAI SDK

export LAOZHANG_API_KEY="YOUR_LAOZHANG_API_KEY"

• Python: run python -m pip install openai.
• Node.js: run npm install openai, save the code as openai-test.mjs, and run it with node openai-test.mjs.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["LAOZHANG_API_KEY"],
    base_url="https://api.laozhang.ai/v1",
    timeout=120.0,
    max_retries=0,
)
response = client.chat.completions.create(
    model="gpt-5.6-luna",
    messages=[{"role": "user", "content": "Reply only with: connected"}],
)
print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.LAOZHANG_API_KEY,
  baseURL: "https://api.laozhang.ai/v1",
  timeout: 120_000,
  maxRetries: 0,
});
const response = await client.chat.completions.create({
  model: "gpt-5.6-luna",
  messages: [{ role: "user", content: "Reply only with: connected" }],
});
console.log(response.choices[0].message.content);

• Text: client.chat.completions.create
• Responses: client.responses.create
• Images: client.images.generate and client.images.edit
• Transcription: client.audio.transcriptions.create
• Speech: client.audio.speech.create
• Embeddings: client.embeddings.create

​

Chat Completions or Responses

• Chat Completions: most OpenAI-format clients use this endpoint, so keep existing messages-based code on it and read the reply from choices[0].message.content. OpenAI sets extra rules for GPT-6 reasoning effort and sampling parameters in Chat Completions; see the GPT-6 launch announcement. Since November 4, 2026, GPT-6 and GPT-5.6 can reason and call tools in the same Chat Completions request on LaoZhang API, so you don’t need Responses just for tools.
• Responses: LaoZhang API supports OpenAI’s Responses API, and every OpenAI text model can use it. OpenAI makes o3-pro available only in the Responses API. Send input, and read output_text in the Python SDK.
• Code already built on Responses: switch the base URL and API key to LaoZhang API and keep the rest. Before you move to a model from another vendor, send a minimal request to confirm it supports Responses; the Responses API examples show one.
• Tools and web search: Function calling covers both endpoints and which models support each. Web search is available only as the Responses web_search tool; see Web search.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["LAOZHANG_API_KEY"],
    base_url="https://api.laozhang.ai/v1",
    timeout=600.0,
    max_retries=0,
)
response = client.responses.create(
    model="o3-pro",
    input="Reply only with: connected",
)
print(response.output_text)

​

GPT-6 and GPT-5.6 pricing

• When one request has more than 272K input tokens, the whole request is billed at a higher tier. For gpt-5.6-sol, that tier is $8 input, $30 output, and $0.8 cache read. The catalog’s tiered pricing table lists every tier.
• The current gpt-5.6-sol price follows OpenAI’s price cut of August 21, 2026, which OpenAI describes as promotional pricing available at least through November 21, 2026 (OpenAI API changelog). LaoZhang API’s pricing configuration adopted the new price on September 10, 2026; the GPT-5.6 Sol price cut announcement compares the old and new rates.
• Prices for GPT-5.4, o3-pro, and the audio and embedding models are in the OpenAI section of the catalog. For per-call and token-based GPT Image billing, see GPT Image 2.5.

​

How this differs from calling OpenAI directly

• API key: create it in LaoZhang API Token management. OpenAI platform keys do not work with LaoZhang API, and a LaoZhang API key does not work with api.openai.com.
• Model IDs: the catalog is the source of truth, and not every model in OpenAI’s catalog is available. Some IDs, such as gpt-image-2.5-flare-vip and gpt-image-2.5-web, are LaoZhang API route IDs that do not exist at OpenAI.
• Token groups: your key’s token group decides which models and routes it can use.
  • The text, audio, and embedding models on this page are in the default group.
  • GPT Image has per-call models such as gpt-image-2.5-flare-vip in default.
  • Official API forwarding runs in the OpenAI Official (formerly Sora2Official) group (same price as OpenAI) and the OpenAI Official Enterprise (formerly GPTImage2 Sora2 Enterprise) group (OpenAI’s price plus 20%), billed by tokens. See GPT Image 2.5.
• Billing: charges come from your LaoZhang API balance at catalog prices, separate from any OpenAI account credit or invoice. Call logs show usage and cost for each request, and Billing, token groups, and invoices explains the rules.
• Other endpoints: if you need an endpoint that is not in the table, such as Batch, Files, or Realtime, check with support at hi@laozhang.ai first.
• Video: OpenAI removed the Videos API, sora-2, and sora-2-pro from its API on September 24, 2026, and LaoZhang API stopped its Sora 2 route on July 1, 2026. For video generation, use Wan 2.7 or Seedance 2.0; see Sora 2 API retired.
• Data handling: requests sent through LaoZhang API are covered by the LaoZhang API Data Policy.

​

Troubleshooting

• 401: check that you are using a LaoZhang API key and that the environment variable is set in the running process.
• 404 or model not available: check the full model ID, the endpoint, and the token group. For example, o3-pro requests must go to Responses. See model availability.
• 400: remove optional fields you added, then correct the request using the error body. Models accept different parameters, so do not copy one model’s settings to another.
• The response arrives but your app shows nothing: parse the structure that matches the endpoint: choices for Chat Completions, output for Responses, and data for Images.
• Client timeout: the request may still be running or may have finished. Check call logs before you resubmit, and do not let both the SDK and your application retry.

​

Related documentation

• Text generation API: protocols and models
• OpenAI protocol: Chat Completions
• API integration guide: Responses, image generation, and image editing examples
• Images API
• Models API
• OpenAI: models, Responses reference, Chat Completions reference, and deprecations