Switch a key's channel, not your code
You choose each API key's channels and can change them in the console at any time, while the key and base URL stay the same.
How to switch channelsStack monthly plans and pay-as-you-go pricing to fit your usage.
curl https://www.su8.codes/v1/chat/completions \ -H "Authorization: Bearer $SU8_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model": "gpt-6.1-sol", "messages": [{"role": "user", "content": "Hello"}]}'
Pick your channel, check every call, keep the protocol your tool already speaks, and handle payment and invoices yourself.
You choose each API key's channels and can change them in the console at any time, while the key and base URL stay the same.
How to switch channelsTime to first token, latency, output speed, token breakdown, and cost are listed for every call from the last 30 days.
Chat Completions, Messages, Responses, and Gemini endpoints all work, and when your tool and the channel use different protocols we convert between them so streaming and tool calls keep working.
Check out with Stripe, then generate a Stripe Invoice in the console or apply for a Chinese tax invoice.
Daily users save with a plan; variable usage pays as it goes. Keep both, and balance takes over once the day's plan quota runs out.
Prices cover one validity period; daily quota is measured at official prices. Unused quota carries over.
Prices are official list prices. Landed discount is what you pay through the cheapest channel; shorter bars cost less.
| Model | Official input | Official output | Landed discount |
|---|---|---|---|
| GPT-6.1 Sol gpt-6.1-sol | $2/ 1M | $10/ 1M | 92.5% offCodex |
| GPT-5.6 Sol gpt-5.6-sol | $4/ 1M | $20/ 1M | 92.5% offCodex |
| GPT-6-astra gpt-6-astra | $10/ 1M | $50/ 1M | 92.5% offCodex |
| GPT-6-sol gpt-6-sol | $2/ 1M | $10/ 1M | 92.5% offCodex |
| Claude Opus 5.5 claude-opus-5-5 | $4/ 1M | $20/ 1M | 77.6% offClaude |
| GPT-5.6 Terra gpt-5.6-terra | $2/ 1M | $12/ 1M | 92.5% offCodex |
It is how much less than the official price you pay. 92.5% off means you pay 7.5% of the official price. Channels differ; the model detail page lists each channel.
Not when the key allows using balance after the plan quota: the next request is charged to balance, and you only see an insufficient quota error when balance is also empty. Plan quota resets daily at 00:00 Beijing time.
Two things: set the Base URL to https://www.su8.codes/v1 and use a key created in the console. OpenAI SDK, Anthropic SDK, Codex, and Claude Code all work directly.
Edit the key under API keys in the console, pick another channel in the Channels step, and save. Within 15 seconds new requests use the new channel; the key and base URL stay the same. Landed discounts can differ by channel and plans cover only some channels, so compare on the model detail page before you switch.
Usage in the console lists time to first token, latency, output speed, token breakdown, and cost for every call, covering the last 30 days.
Yes. Daily quota and RPM add up across plans; buying the same plan again extends its validity.
Two kinds. A Stripe Invoice has no service fee: generate and download one for a paid plan or top-up order under Billing > Orders in the console, except plan orders bought with platform invoicing. For a Chinese tax invoice, choose platform invoicing when you buy a plan or apply later for existing orders; it carries a 5% service fee and a minimum invoice amount of ¥300.
Balance never expires and works with every enabled model.