GPT-5.6 Sol API relay: 59API balance or monthly card?
The problem: a subscription bill you did not ask for
Most developers meet the same awkward choice when they start using an AI API. A monthly subscription feels predictable on paper, but the bill arrives whether or not you shipped anything, wrote a prompt, or opened the dashboard. The alternative is pay-as-you-go, which removes the fixed charge but leaves you watching a balance and estimating usage. Neither model fits every team, and every month you guess wrong costs time or money.
59API was built around the pay-as-you-go side of that trade-off. It is an independent third-party AI API relay: one account and one API key, one API host at https://api.59api.com, and access to Claude, GPT, Gemini, DeepSeek and other models. You top up a prepaid balance, and each request is charged according to real token usage. There is no monthly fee, and the balance never expires. Monthly and weekly cards exist too, but they are optional rather than required.
- Want to start small and pay only for what you call? Use the balance.
- Want a predictable ceiling and a fixed window of access? Consider a card.
- Want both? Keep a balance for exploration and add a card for a focused burst of work.
The rest of this checklist breaks down the numbers, the rules that catch people out, and the setup steps, so you can pick a payment path with your eyes open.
How the prepaid balance actually works
The balance model is simple enough to explain in one sentence: you buy credit, and requests consume it based on the tokens they actually use. That means input tokens, output tokens and cache reads are all metered, and rates are quoted per 1 million tokens. Billing is in CNY, with USD figures shown at ¥7.2 = $1, which makes it straightforward to compare a token rate against a model provider list price.
Top-ups come in six sizes: ¥10, ¥50, ¥100, ¥200, ¥500 and ¥1000. You can pay by WeChat Pay or bank card, and invoices are available. Because there is no monthly fee, a small top-up is enough to test an integration without committing to a recurring charge. And because the balance never expires, unused credit stays with you, which is a meaningful difference from a subscription period that resets.
- Start with a small top-up and watch real usage.
- Add more only when your traffic pattern becomes clearer.
- Top up again whenever you want; there is no renewal deadline.
One important detail: the balance key is the default path into 59API. If you later add a card, the two do not mix. A card key calls only that family of models and is paid only from the card, never falling back to wallet balance. A balance key never draws on a card either. Keeping that separation in mind makes the decision much easier to reason about.
What balance usage looks like in real numbers
Rates tell you little until you attach them to a workload. Take GPT-5.6 Sol, model id gpt-5.6-sol. On 59API, input is $0.3472 (¥2) per 1 million tokens, output is $1.94 (¥14), and cache read is $0.0278 (¥0.2). The official list price reference is $4 / $20. Now imagine a workload of 2M input tokens plus 0.5M output tokens without cache: that comes to $1.67 (¥12) on 59API, against $18 (about ¥129.6) at official list price, so 59API is about 9% of list price.
Claude pricing follows the same per-token logic. Claude Sonnet 5.5, model id claude-sonnet-5-5, lists input at $1.11 (¥8), output at $5.56 (¥40) and cache read at $0.0972 (¥0.7) per 1 million tokens on 59API. For 3M input plus 0.5M output tokens without cache, the cost is $6.11 (¥44) on 59API, compared with $11 (about ¥79.2) at official list price, so 59API is about 56% of list price.
- Estimate monthly input and output tokens separately; output is usually the larger line item.
- Check your cache read volume, since it carries its own rate.
- Re-estimate when your prompt structure or model choice changes.
These are worked examples, not quotes. Your own mix of tokens is what determines your bill, and the balance lets you discover it without a fixed commitment.
What the optional cards add, and what they do not
Cards are for developers who want a defined spending window. Each card has its own key, can call only that family of models, and draws only on its own quota. There is no fallback to your wallet balance, and unused quota does not roll over. Quota and caps are card metering units written in $, not official list-price dollars, so they should not be converted into official prices.
For Claude, C-PRO Standard covers 30 days for $152.64 (¥1099), with quota $2650 and a weekly cap $1000, draining at the card metering price × 2.4. Its reference value at official list prices is about $736.11 (¥5300). For a shorter window, the Weekly G-PRO $1500 card covers GPT for 7 days at $40.56 (¥292), with quota $1500 and a weekly cap $1500, also draining at the card metering price × 2.4.
- A card fixes the period, the family of models and the quota.
- It does not merge with your balance, and leftovers are gone when the window ends.
- Weekly cards suit a concentrated sprint; 30-day cards suit a steady month.
That separation is a feature, not a trap. If you want a card for a Claude-heavy week and keep a balance for GPT experiments, the two live side by side without one quietly consuming the other.
Your decision checklist and the first request
Ask three questions before you pick. First, is your usage steady or spiky? Steady usage makes a card easier to reason about; irregular usage makes a balance more natural. Second, do you want a hard ceiling? A card quota gives you one. Third, do you need more than one model family at the same time? If yes, the balance is the simpler common denominator.
Then connect. For Codex, OpenAI SDKs, cURL and OpenAI-compatible clients such as Cursor, Cherry Studio, Cline and OpenCode, use base_url https://api.59api.com/v1 and chat endpoint https://api.59api.com/v1/chat/completions; these clients pair with GPT, DeepSeek and other non-Claude, non-Gemini models only. For Claude Code or Claude Desktop, set ANTHROPIC_BASE_URL=https://api.59api.com and ANTHROPIC_AUTH_TOKEN=YOUR_59API_KEY, and use Claude Code 2.1.285 or newer. For Gemini CLI, set GOOGLE_GEMINI_BASE_URL=https://api.59api.com and GEMINI_API_KEY=YOUR_59API_KEY; Gemini models use Google native protocol only and do not work on /v1/chat/completions.
- Top up the smallest amount that fits your test, send a real request, and read the usage.
- Upgrade to a card when you can name the window and the model family you need.
- Keep the balance for everything else; 59API never expires it.
That is the whole pitch: flexible payment on one account, one key and one host, with per-request pricing you can inspect and optional cards when a fixed window suits the job.