Skip to main content
LaoZhang API (laozhang.ai) is an AI API gateway operated since 2025 by YingTu Technology Pte. Ltd. in Singapore. One API key reaches Google Nano Banana, OpenAI GPT Image, and xAI Grok Imagine image models, plus text and video models. This page collects the facts to check before you integrate, last verified on October 10, 2026. If this page conflicts with the formal terms, the terms control.

​
Model sources and routes

Every model on LaoZhang API comes from the vendor’s official channel; there are no self-trained or substitute models. Models are served on two kinds of routes, and the token group you choose when creating a key decides which one a request uses:
  • Official API forwarding (official-forward): the request is forwarded to the vendor’s official API and billed by the vendor’s token rules. For GPT Image, the GPTImage2 Sora2 Enterprise group uses OpenAI keys only, and Sora2Official mixes Azure OpenAI and OpenAI keys; see GPT Image 2.5.
  • Per-call routes: the same official models, served by LaoZhang API at a fixed price per call or per returned image that is lower than token billing on the official-forward route. The two routes are independent. Nano Banana, Grok Imagine, and the GPT Image -vip and -web models are per-call models; supported parameters, sizes, and editing follow each model guide.
The same model can have different model IDs, parameter ranges, and prices on the two routes, so check the model guide before you copy parameters from one route to the other. The current status, price, and groups of every model ID are listed in Models and pricing and in the console. Launches, retirements, and price changes are posted in Announcements; past announcements are not rewritten.

​
Billing, payment, and refunds

Image models follow the billing rules in their guides: Nano Banana charges once when the upstream returns HTTP 200, even without an image; Grok Imagine charges per successfully returned image; GPT Image requests that return an error without generating an image are generally not charged. See Nano Banana billing, Grok Imagine, and GPT Image 2.5.

​
Data handling

  • API prompts and responses are not stored by default. They are transmitted only as needed to process the request and return the result, and are not used for model training or profiling by default.
  • Necessary request metadata is kept for billing, usage display, troubleshooting, rate limiting, and security.
  • Upstream model and infrastructure providers may apply their own logging and retention rules.
How LaoZhang API handles data and logs and the Data Policy give the full boundaries.

​
Service status and change notices

  • Status page: live availability of each endpoint.
  • Announcements: launches, price changes, suspensions, and recoveries, archived by date.
  • Telegram channel @laozhang_ai: the same notices as the announcements page.
The Terms do not promise a fixed service level. For production traffic that needs the highest stability, use the enterprise official-forward group, which runs on official keys only, and implement timeouts and retries in your client.

​
Getting access

​
Support

When you report a problem, include your account email, the model ID, the request time, and the redacted error. Never send a full API key or payment evidence.

​
FAQ

​
How long has LaoZhang API been operating?

Since 2025, under YingTu Technology Pte. Ltd. Model changes and service status over time are recorded in the announcement archive.

​
Are these the vendors’ original models?

Yes. Both the official-forward and the per-call routes use the vendor’s official channel; they differ in billing. Official forwarding charges the vendor’s token rates, and per-call routes charge a fixed price that is usually lower. The two routes use different model IDs, and each model guide states which route a model ID belongs to, with its parameter range and price.

​
Can I use these image models without a vendor account?

Yes. Your LaoZhang API key is the only credential, and billing happens only on LaoZhang API. You still have to follow each vendor’s usage policies for the generated output.

​
Am I charged when no image comes back?

It depends on the model: Nano Banana charges when the upstream returns HTTP 200, GPT Image generally does not charge requests that produce no image, and Grok Imagine charges per successfully returned image. The call log is authoritative; the rules are in each model guide.