For reviewers of comparison / review / directory sites: self-serve verification of TokenDos (an API relay service for developers based on shared account hosting). All links are public. Updated on 2026-10-03.
| Site | TokenDos · https://www.tokendos.com/tokendos |
|---|---|
| API base | https://api.tokendos.com (OpenAI /v1 · Anthropic /anthropic) |
| Launched | 2026-06 |
| Operator | 个人站长（无公司主体） |
| Top-up ratio | 1:1; 首充可领取 10% 代金券（充值时自愿选用，可与召回券叠加）。 |
| Minimum top-up | 1 USD |
| Balance validity | 无有效期与清零周期，按 token 用量实时扣减。 |
| Refund policy | 余额按用量实时扣减；如出现重复扣费、计费异常或未消耗余额退款需求，请通过官网 /support、QQ 售后群或 business@tokendos.com 提交账号邮箱、订单号与交易日期，经售后核查后处理（PayPal 订单支持原路退款）。 |
| Support | 官网 /support 页面；QQ 售后群；商务邮箱 business@tokendos.com。 · /support |
| Contact | business@tokendos.com |
Top 15 cheapest listed models by live rate. Rate follows the industry convention used by comparison sites: rate = site price in CNY / official catalog price = (site price in USD * FX rate) / catalog price (site prices are quoted in USD while ranking sites compare in CNY), so figures stay comparable across sites. A rate is a single value (site input price vs official input price; 1.0x = at official price). Full data: /llms-full.txt and /api/tokendos/public/transit-snapshot.
| Model | Site input /1M | Official input /1M | Rate (x) |
|---|---|---|---|
gpt-5.6-luna | $0.0001 | $1 | 0.000671x |
glm-5.3-flash | $0.0001 | ¥0.8 | 0.00125x |
deepseek-v4.1-flash | $0.0006 | ¥2 | 0.002x |
Qwen3.8-Flash-Next | $0.0149 | ¥12 | 0.00832x |
deepseek-v4-flash | $0.0004 | $0.149 | 0.018x |
gpt-5.6-terra | $0.0089 | $2.5 | 0.024x |
gpt-5.5 | $0.0223 | $5 | 0.03x |
gpt-5.6-sol | $0.0223 | $5 | 0.03x |
gpt-6.1-sol | $0.0089 | $2 | 0.03x |
gpt-6-astra | $0.0447 | $10 | 0.03x |
gpt-6-sol | $0.0089 | $2 | 0.03x |
grok-4.7 | $0.0144 | $2 | 0.0483x |
gpt-6-luna | $0.0007 | $0.1 | 0.05x |
grok-4.6 | $0.02 | $2 | 0.0671x |
grok-4.5 | $0.02 | $2 | 0.0671x |
| Requirement | Evidence / notes |
|---|---|
| 1. Accessible website | https://www.tokendos.com/tokendos |
| 2. Public pricing page / read-only API | https://www.tokendos.com/tokendos-market（live lowest prices） · https://www.tokendos.com/tokendos-pricing · https://www.tokendos.com/api/tokendos/public/terms |
| 3. Public top-up ratio | 1:1 (1 USD = 1 USD balance; +10% voucher on first top-up) · https://www.tokendos.com/tokendos-pricing |
| 4. Minimum top-up / validity / refund / support | https://www.tokendos.com/tokendos-pricing · https://www.tokendos.com/support |
| 5. Model group rates | Marketplace model: no fixed group multipliers; rate = live lowest price ÷ official price, published at https://www.tokendos.com/tokendos-pricing and /llms-full.txt; machine-readable at terms (modelRates) |
| 6. Upstream / key pool type | 混池：供应方自有官方 API Key + 订阅账号转 API（OAuth），平台审核资质并按真实调用动态监控 · 供应方自有 Key/账号池聚合（云端托管 + 本地托管双模式） · snapshot disclosure field |
| 7. Availability monitoring (per model & per group) | https://www.tokendos.com/tokendos-pricing（7-day per model and per protocol group） · https://www.tokendos.com/api/tokendos/public/availability · https://www.tokendos.com/api/tokendos/public/transit-snapshot（monitoring / group_monitoring） |
| 8. Comparable official prices | Official input/output prices at https://www.tokendos.com/tokendos-pricing and terms (modelRates with source) |
| 9. Rate limits | 消费侧熔断：同一 API Key 在 60 秒窗口内失败请求达到 30 次即临时拒绝 180 秒（返回 429）；上游节点按响应码动态退避与故障转移，账号级默认并发 1。未声明额外 RPM/TPM 配额，实际吞吐以上游返回为准。 |
| Endpoint | Content |
|---|---|
https://www.tokendos.com/api/tokendos/public/terms | Terms: top-up ratio, minimum top-up, refund, group disclosure, official price table (modelRates) |
https://www.tokendos.com/api/tokendos/public/availability | 7-day availability: calls & success rate per model and per protocol group (real call logs) |
https://www.tokendos.com/api/tokendos/public/transit-snapshot | ai_transit_snapshot v1: groups + monitoring + group_monitoring + billing + disclosure |
https://www.tokendos.com/llms.txt | Site notes for AI crawlers (llms.txt standard) |
https://www.tokendos.com/llms-full.txt | Full live price table (Markdown, all listed models) |
Why no fixed group multipliers? TokenDos is a marketplace: any key holder can list, and the site shows the live lowest price across nodes. The rate (= lowest price ÷ official price) is computed in real time and published for verification.
What does 'mixed pool' mean? 混池：供应方自有官方 API Key + 订阅账号转 API（OAuth），平台审核资质并按真实调用动态监控
How is availability computed? Aggregated from real proxy call logs per UTC day; success = HTTP-layer success. Only currently listed models are counted. Groups refer to protocol groups (OpenAI /v1, Anthropic /anthropic).
How fresh is the data? Pricing pages are live; APIs cache for up to 5 minutes; availability aggregates daily and stays consistent between pages and APIs within the cache window.