Buying API access for GPT, Claude, Gemini or Kimi starts with the model and protocol your application needs. Xinghuo AI Gateway provides a shared entry point for models authorized for your key. Check the current catalog and your account quote before choosing credits.
Four checks before checkout
| Check | Where to look | What to confirm |
|---|---|---|
| Model and protocol | Model catalog and integration guides | Exact model ID; Chat Completions, Responses or Messages support |
| Effective price | Account pricing and usage details | Input, output, cache charges and any group multiplier |
| Credits and expiry | Product, order and account | Balance units, validity, plan limits and key limits |
| Payment and support | Checkout and site support | Available payment methods, invoice and refund terms |
From account to your first request
- Browse the current models and prices.
- Create an account or sign in and confirm the quote shown for your account.
- Open Buy credits to see the available products and payment methods. If you already have a code, use account redemption.
- Create a site-issued API key and configure the protocol, base URL and model.
- Make a short request with a limited output, then verify its cost in usage lookup.
Does the amount paid equal the API balance?
Do not assume that it does. Payment currency, balance units, bonuses, group rates and key limits can differ. Use the checkout terms and credited balance. See token billing for cost calculations.
Is there a free trial?
Registration credits, trials and promotions depend on what the live registration page and your account actually show. This guide does not promise a fixed free allowance or a permanent free plan.
Can these credits replace a ChatGPT or Claude web subscription?
No. Site-issued API credits are for API requests and compatible clients. They do not grant a provider web subscription. Read API credits versus web subscriptions before purchasing.
Evaluate your actual workflow
Check response quality, tools, streaming, latency and cost with your own tasks. Keep model ID, inputs, output limits and concurrency consistent when comparing results. Record the test time. For support, share an order or request ID and redacted details, never your complete API key.