This DeepSeek API key comes from a third-party gateway provider, not directly from DeepSeek. Each pack gives you a token allowance to call DeepSeek V4 Pro and V4.1 Flash through an OpenAI-standard base URL, so Cursor, VS Code, Codex, OpenCode or a Python script accept it right away. Tokens never expire, and you pay with crypto instead of topping up a DeepSeek account.
What's included
- An API key, a base URL, a page showing remaining tokens and setup instructions
- DeepSeek V4.1 Flash: for everyday chat, writing, code fixes and reading screenshots; the only model here that accepts images
- DeepSeek V4 Pro: for harder problems such as agent planning, multi-file debugging and long reasoning
- Both models offer a 1M-token context, up to 384K output tokens, tool calling, JSON output and thinking mode on by default
Which plan should I choose?
Usage is multiplied per model: V4 Pro counts at ×0.5 and V4.1 Flash at ×0.3. So one pack gives you about twice its size on V4 Pro and more than three times on Flash.
- 10M tokens (about 20M on Pro / 33M on Flash): try Flash in Cursor for a few Q&A and bug-fix sessions.
- 25M tokens (about 50M / 83M): a small project with V4 Pro planning and editing many files.
- 50M tokens (about 100M / 167M): daily coding with Cline or Cursor for several weeks.
- 100M tokens (about 200M / 333M): large repos or batch translation and summarization on Flash.
- 500M tokens (about 1B / 1.67B): multiple repos or large data pipelines.
- 1B tokens (about 2B / 3.3B): long-running agents and very high text volume.
Every pack includes both models. Larger packs cost less per million tokens.
Who it's for
- Developers who want a strong, low-cost coding model in their IDE
- Builders running batch translation, summarization or classification
- Anyone who wants tokens that don't expire
How it works
- Choose a token pack.
- Pay with crypto via Cryptomus (USDT, USDC, BTC, ETH, LTC, TRX and more). Your order is confirmed automatically on-chain.
- Receive your key, base URL and setup guide by email, usually within 30 minutes during support hours (8:00–22:00 GMT+7). Setup help is available on Telegram.
Important notes
- Third-party key, not an original DeepSeek key: it cannot log in to DeepSeek's platform or call DeepSeek's API endpoint directly. Do not send sensitive data through it.
- No expiry: use tokens until they run out, then buy another pack.
- Input and output tokens are multiplied by the model factor (V4 Pro ×0.5, V4.1 Flash ×0.3). There is no prompt caching, so each turn counts the full context again.
- Warranty: if the key fails, we replace it or refund the unused tokens. Not covered: used-up tokens, misconfiguration, exceeding request limits or sharing the key.
