A unified API gateway compatible with OpenAI / Anthropic / Gemini protocols, billed by usage.
Multiple independent supply channels are aggregated; requests are routed by realtime scoring, health probing and automatic failover.
Available models, prices and specs are as shown live in the model catalog; prices adjust automatically with supply-side cost changes.
2. Accounts and tokens
Provide truthful registration information; you are responsible for all activity under your account.
API tokens are your credentials — keep them safe (never embed them in frontends or public repos). Consumption from a leaked token is borne by the account holder.
To protect the platform and its users, we may rate-limit, suspend or ban accounts for abusive usage (volume fraud, token theft, attacks, resale abuse).
3. Billing and top-ups
Usage-based: text models bill per token; image/video models bill per call, as labeled in the catalog.
Never billed on failure: failed requests cost nothing and pre-deducted quota is refunded automatically; only successful calls settle.
Top-ups: Alipay and USDT and other channels; funds appear in your wallet and apply to all services.
Refunds: unconsumed balance can be refunded via ticket (returned by the original payment route, minus payment-channel fees); consumed credit is non-refundable.
4. Acceptable use
No illegal use, including generating or distributing unlawful content.
No attacks, scanning, rate-limit evasion or resource abuse against the platform or suppliers.
No reselling or lending accounts/tokens for fraud, money laundering or other unlawful purposes.
No using model output to infringe others' IP, privacy or reputation.
Violations may lead to immediate suspension; unconsumed balance is handled per the severity of the violation.
5. Availability and disclaimers
We keep multi-supplier failover best-effort, but do not guarantee the permanent availability of any single model: suppliers may throttle, delist or change models.
For outages caused by force majeure, upstream failures, network faults or maintenance, we restore service as fast as we can; impact is handled under the "unconsumed balance is refundable" principle.
Model output is generated by the respective suppliers; we do not warrant its accuracy or compliance. Do not use output directly for high-stakes medical, legal or financial decisions.
6. Intellectual property
The platform's interface, documentation and self-developed features belong to 321AI. Model names and trademarks belong to their respective owners; we provide relay and scheduling services only.
7. Changes to these terms
We may update these terms; updates are published on this page. Continued use constitutes acceptance; if you disagree, stop using the service and request a refund of unconsumed balance.