Skip to main content
LaoZhang API charges requests against a prepaid balance. Each model is billed by usage (tokens) or per call, and the group of the API key you use can apply a price multiplier.

​
Usage-based and per-call billing

Text, embedding, moderation, and audio models are usage-based. Most image models, including Nano Banana, FLUX.2, Seedream, and the GPT Image 2.5 VIP models, are billed per call. Video models follow the billing rules in the Wan 2.7 and Seedance 2.0 guides. Some models are priced both ways, depending on the group:
  • gpt-image-2 is billed per call in the default group and by tokens in the official API forwarding groups, OpenAI Official (formerly Sora2Official) and OpenAI Official Enterprise (formerly GPTImage2 Sora2 Enterprise).
  • The official-forward GPT Image 2.5 models, gpt-image-2.5-flare and gpt-image-2.5-sunburst, are also billed by tokens at the same rates.
  • GPT Image 2.6, gpt-image-2.6-flare and gpt-image-2.6-sunburst, launched on November 5, 2026 with the same token rates as 2.5 in the official-forward groups. For SVG output, writing the SVG is free with the default gpt-6.1-terra and eight other low-cost text models; other writing models are billed separately at their official rate. The default group also has per-call versions at $0.03:
    • gpt-image-2.6-flare-vip, gpt-image-2.6-sunburst-vip
    • gpt-image-2.6-web, which can return animated GIFs
Since November 6, 2026, official model names can also be billed per call. They run on the VIP route and behave exactly like their -vip counterparts, at $0.03 per image with native 2K/4K:
  • In default: the four official 2.5 and 2.6 model names above.
  • In GPT Image VIP: the same four names, plus gpt-image-2.
  • If your code or tool only accepts official model names, switch to a GPT Image VIP token and keep model as it is.
The per-call web version route of GPT Image 2.5 uses the model ID gpt-image-2.5-web. The older gpt-image-2-web is a legacy ID kept for existing integrations and also supports generation and editing. It is a separate route, so do not reuse its parameters with gpt-image-2.5-web. Token counts depend on the model’s tokenizer, the conversation history and files you send, tool calls, and output length. Instead of estimating from word counts, read the input tokens, output tokens, and charge of each request in call logs. Some models also use tiered prices that depend on how many tokens a single request uses.

​
Token billing modes

Every API key, called a token in the console, has a billing mode. It decides which models the token can call and which price applies. You set it when you create or edit the token in Token management; Usage first is the right choice in most cases: A token needs both the right billing mode and the right group. These models work only in their own groups, each with Usage first or Usage-based billing:
  • Wan only in the Wan group.
  • Seedance only in SeeDance2.
  • Official-forward GPT Image only in its official-forward groups.
You rarely need the console’s other options. A few models have their own requirement, such as a Per-call token for per-call gpt-image-2 in the default group; follow the model guide in those cases. To confirm which billing method a request used, find it in call logs.

​
Token groups and multipliers

A token’s group decides which routes and models it can use. Some groups apply a multiplier to the listed model price: requests in the group are charged the listed price multiplied by that value. For example, OpenAI Official Enterprise has a 1.2× multiplier. The group list in token settings shows each group’s multiplier. Per-call GPT Image models such as gpt-image-2-vip and gpt-image-2.5-web are also offered in another image group (see the console for its name). Models and pricing lists each model’s groups and prices, with any multiplier after the group name. Confirm the prices for your account in the console.

​
Check prices and charges

  • Models and pricing: list prices, billing method, token groups, and endpoints for each model, with the time the data was last updated.
  • Console pricing: the prices and groups that apply to your signed-in account.
  • Call logs: the model, token usage, and charge of each request.
  • Balance query API: remaining and used quota for alerts and dashboards.
To estimate a budget, send a few representative requests, read their usage and charges in call logs, and scale by your expected volume. An upstream provider’s public price does not predict your LaoZhang API bill, except where a model guide gives the conversion, as the Wan 2.7 and Seedance guides do. To cap spending per service, give each service its own token with a quota limit.

​
Top up and pay

Add funds to your balance from the top-up page in the console. The top-up page takes card payments through Stripe, and businesses can pay by bank transfer through support; LaoZhang API does not store full card numbers. Purchases and refunds are governed by the Terms, and the payments and refunds FAQ summarizes them.

​
Enterprise purchasing and invoices

For enterprise payments, contracts, invoices, or volume pricing, contact support by email at hi@laozhang.ai or on Telegram at @laozhang_cn. Include:
  • Your company name and account email.
  • The models you plan to use and your estimated monthly usage.
  • Your contract and invoice requirements.
Support confirms the terms and invoice details for your account. LaoZhang API is operated by YingTu Technology Pte. Ltd., a Singapore company.

​
FAQ

​
Am I charged when a request fails?

Non-200 upstream responses aren’t charged, including invalid parameters, authentication failures, 429 rate limits, and upstream 5xx errors. After an HTTP 200, usage-based models bill actual tokens and per-call models bill per call or per returned image; Nano Banana charges an HTTP 200 even when no image comes back. A client timeout isn’t a failure, so check call logs before you resend. The image-model rules are in Billing and failure rules.

​
Why is a charge higher than I expected?

Open the request in call logs. Common reasons are a group multiplier, more input tokens than expected (for example, a long conversation history or attached images), a tiered price for a large request, or a model that your group bills by tokens rather than per call. To convert the ratios in a log entry’s billing details into prices, see Read the charges in a log entry.

​
Does a request that timed out cost anything?

A client timeout does not cancel the request on the server, and it does not tell you whether a charge occurred. Check call logs before you retry, especially for image and video requests.

​
Can I get a volume discount?

The documentation does not publish fixed discount tiers. Contact support for enterprise or volume terms.