DaoXEWhat you actually pay for#
With official APIs you pay each vendor's list price per 1M input/output tokens, plus the hidden overhead of running multiple accounts: separate cards, separate invoices, and — in many regions — failed checkouts. A gateway changes the shape of the bill: one key, one balance, per-model pricing, and the option to route cheaper models for cheap tasks.
Gateway vs official: the honest comparison#
| Dimension | Official APIs (direct) | DaoXE gateway |
|---|---|---|
| Accounts & billing | One per vendor | One key, one bill |
| Checkout in your region | Sometimes fails | Overseas-ready (not mainland China) |
| Pricing model | Per-vendor list price | Per-model; default at parity |
| Price vs official list | Baseline | Lower via low-price groups — see the real table below |
| Model coverage | One vendor per key | GPT · Claude · Gemini · DeepSeek · Kimi · Qwen · Doubao |
| Claude protocol | Native | Native /v1/messages |
| Integrity check | Trust the vendor | Open, reproducible benchmark |
Real price comparison (crawled 2026-07-19)#
Real DaoXE prices vs each provider's official list price, USD per 1M tokens (input / output). DaoXE's default group sits at official parity — the savings come from actively choosing a cheaper group, not from the default. Numbers drift and are account-scoped; the live pricing page is authoritative.
| Model | DaoXE default | Official list | Low-price group (example) | That group | vs official |
|---|---|---|---|---|---|
gemini-2.5-pro | 1.25 / 10 | 1.25 / 10 | Ant-Gemini (0.30) | 0.375 / 3.0 | −70% |
gemini-3.5-flash | 1.50 / 9 | 1.50 / 9 | Ant-Gemini (0.30) | 0.45 / 2.7 | −70% |
claude-opus-4-5 | 5 / 25 | 5 / 25 | claude-官方max-满血 (0.70) | 3.5 / 17.5 | −30% |
claude-sonnet-4-5 | 3 / 15 | 3 / 15 | claude-官方max-满血 (0.70) | 2.1 / 10.5 | −30% |
claude-haiku-4-5 | 1 / 5 | 1 / 5 | claude-kiro-逆向 (0.15) * | 0.15 / 0.75 | −85% * |
gpt-5.5 | 5 / 30 | 5 / 30 | codex-pro (0.20) | 1 / 6 | −80% |
gpt-5.4 | 2.5 / 15 | 2.5 / 15 | codex-pro (0.20) | 0.5 / 3 | −80% |
grok-4.5 | 2 / 6 | 2 / 6 | grok-福利 (0.05) * | 0.10 / 0.30 | −95% * |
grok-4.3 | 1.25 / 2.5 | 1.25 / 2.5 | grok-福利 (0.05) * | 0.062 / 0.125 | −95% * |
deepseek-v4-pro ⚠ | 1.74 / 3.48 | 0.435 / 0.87 | 海外国产模型 (0.50) | 0.87 / 1.74 | ⚠ +2–4× |
deepseek-v4-flash | 0.14 / 0.28 | 0.14 / 0.28 | 海外国产模型 (0.50) | 0.07 / 0.14 | ≈−50% |
Read this honestly
① The default group ≈ official list price. ② Most Western models (GPT / Claude / Gemini / Grok) via low-price / relay groups run ≈30–80% below official; the most aggressive reverse / subsidy / “福利” channels go lower, but DaoXE itself labels them “quality not guaranteed / medium stability / subsidized” — a stability trade-off, not the headline. ③ Honest counter-example: cheap domestic models (DeepSeek / Qwen / GLM) have no price advantage here and can cost more than going direct (DeepSeek V4 Pro ≈ 2–4× official). Don't claim “everything is cheaper.” * grok-福利 needs a paid “1.5 元 大乐购” activation first.
Payment & top-up: what you actually hand over#
DaoXE bills each model at its official USD price, but you top up in your own currency at a flat 1 RMB = $1 of credit — across every method, USDT included. At a real RMB/USD rate near 7 (it moves), that top-up is the real saving layered on top of the price table above.
| Method | Accepted | Top-up rate |
|---|---|---|
| Alipay (支付宝) | Yes | 1 RMB = $1 credit |
| WeChat Pay (微信) | Yes | 1 RMB = $1 credit |
| USDT (crypto) | Yes | converted to same-day RMB, then 1 : 1 — so USD holders abroad get the same rate |
Read the saving honestly
Because top-up is 1 RMB = $1 while the real rate is ~7, Western flagships (GPT / Claude / Gemini / Grok) land around 1/7 of official USD cost (≈85% off), and low-price groups go lower still. Cheaper domestic models (DeepSeek / Qwen / GLM) save less: their USD list is marked up here, but the 1:1 top-up still comes in below going direct (roughly −44% on default, ~−72% on a low-price group). Don't read it as “everything is 1/7”. Rates fluctuate; daoxe.com/pricing is authoritative.
When a gateway saves money — and when it doesn't#
Saves you money when…
you use several vendors, route cheap tasks to cheap models, or hit payment/region friction going direct.
Neutral when…
you use exactly one model at high volume and already have working official billing — then compare per-token rates directly.
Only worth it if…
you can verify integrity. Cheap + unverifiable is a false economy.
How the billing stays transparent#
- Per-model, per-token pricing published on the pricing page.
- Account-scoped model list, so the models you see are the ones you can call.
- Independent verification with the open benchmark — you don't have to take the numbers on faith.
Numbers you should confirm yourself
Real per-model prices are in the table above (crawled 2026-07-19); free credit to be confirmed and uptime measure it yourself are still set by the service and measured by you — see daoxe.com/pricing.
Verify the cost advantage in one run#
The fair way to settle it: run the same fixed prompt through DaoXE and the official API at temperature=0, and compare token usage, latency and output. The benchmark guide automates that.
curl --fail-with-body --show-error --silent \
https://daoxe.com/v1/chat/completions \
-H "Authorization: Bearer ${DAOXE_API_KEY}" \
-H "Content-Type: application/json" \
-d '{
"model": "YOUR_EXACT_MODEL_ID",
"max_tokens": 64,
"messages": [{"role": "user", "content": "Say hello in one sentence."}]
}'Frequently asked questions#
Is DaoXE cheaper than official APIs?
The default group is at official parity; most Western models via low-price groups run ~30–80% cheaper (see the dated table above). Honest caveat: cheap domestic models like DeepSeek can cost more here than going direct. Verify current per-model rates on the pricing page.
Are these exact prices guaranteed?
No — prices drift and are account-scoped. The table above is a dated snapshot (crawled 2026-07-19); the live pricing page is always authoritative.
How is billing metered?
Per model, per token (input/output), on one balance. Your account's usage view and the pricing page are authoritative.
Can I trust the savings aren't a downgraded model?
Verify it: run the open benchmark against DaoXE and the official API and compare.
Try DaoXE — and benchmark it yourself
One key for GPT, Claude, Gemini, DeepSeek and more. Point the open benchmark at us and compare — don't take our word for it.