# bestokens

> OpenAI-compatible AI API gateway. One base URL and one key for GPT, Claude, Gemini, DeepSeek and 100+ models. Pay-as-you-go credits, no monthly fee, no markup on inference.

bestokens is an independent, OpenAI-compatible API gateway operated on Cloudflare's global network. Developers buy prepaid credits and call GPT, Claude, Gemini, DeepSeek, Grok, Llama, Mistral, Qwen, Kimi and GLM models through a single base URL and a single API key. Inference is billed at upstream list price with no markup; revenue comes from a 4%–2.5% platform fee charged at top-up.

## Key facts

- Base URL: https://api.bestokens.cn/v1
- Protocol: OpenAI-compatible (`/chat/completions`, `/models`, `/embeddings`)
- Models available: 100+ across 10+ providers; 20 documented on the public model list
- Pricing model: prepaid credits, no subscription, no per-seat fee, 0% markup on tokens
- Platform fee: 4%–2.5% on top-up (2.5% above $2,000); crypto adds 1%
- Minimum top-up: $5; credits never expire
- Payment methods: card and wallets via Creem (merchant of record), crypto via NOWPayments (BTC, ETH, USDT, USDC, 300+ assets)
- Streaming: server-sent events with a final usage frame when `stream_options.include_usage` is true
- Billing granularity: integer micro-dollars (1 USD = 1,000,000), rounded up per request
- Prompt and completion content is not stored; only token counts, latency and status are retained for 30 days
- Languages supported on the website: 12 (en, zh, ja, ko, es, fr, de, pt, ar, ru, hi, ms), English default

## Popular models

- GPT-5 (`gpt-5`) — OpenAI, 400K context
- GPT-5 mini (`gpt-5-mini`) — OpenAI, 400K context
- Claude Sonnet 4.5 (`claude-sonnet-4-5`) — Anthropic, 200K context
- Claude Haiku 4.5 (`claude-haiku-4-5`) — Anthropic, 200K context
- Gemini 2.5 Pro (`gemini-2-5-pro`) — Google, 1049K context
- DeepSeek V3.2 (`deepseek-chat`) — DeepSeek, 128K context
- DeepSeek R1 (`deepseek-reasoner`) — DeepSeek, 128K context

Full model list and live prices: https://www.bestokens.cn/models

## Pricing

| Top-up | Platform fee |
| --- | --- |
| $5 – $99 | 4% |
| $100 – $499 | 3.5% |
| $500 – $1,999 | 3% |
| $2,000+ | 2.5% |

Pricing page: https://www.bestokens.cn/pricing

## Documentation

- [quickstart](https://www.bestokens.cn/docs/quickstart)
- [authentication](https://www.bestokens.cn/docs/authentication)
- [chat-completions](https://www.bestokens.cn/docs/chat-completions)
- [streaming](https://www.bestokens.cn/docs/streaming)
- [errors](https://www.bestokens.cn/docs/errors)
- [rate-limits](https://www.bestokens.cn/docs/rate-limits)
- [billing](https://www.bestokens.cn/docs/billing)
- [keys](https://www.bestokens.cn/docs/keys)
- [faq](https://www.bestokens.cn/docs/faq)

## Comparison

- vs. calling each vendor directly: same request format, but one key, one bill, one usage log, and automatic failover across upstream channels. No separate accounts or cards per vendor.
- vs. OpenRouter: OpenRouter charges a 5.5% (Standard) or 8% (Business) top-up fee. bestokens charges 4%–2.5% and does not add margin on inference.
- vs. self-hosting: no GPU operations, no capacity planning; pay per token instead of per hour.

## Policies

- Terms of Service: https://www.bestokens.cn/tos
- Privacy Policy: https://www.bestokens.cn/privacy
- Acceptable Use Policy: https://www.bestokens.cn/aup
- Refund Policy: https://www.bestokens.cn/refund
- Service is not offered to residents of mainland China and is blocked at the edge by geolocation.

## Contact

- Support: support@bestokens.cn
- Sales and invoicing: sales@bestokens.cn
- Abuse reports: abuse@bestokens.cn

Last updated: 2026-10-27