DeepSeek V4 Pro + Flash — API Key, $1 Balance, No Subscription
Seller: Anand Ai
Price: 1.10 USD
Available. The final price depends on the selected options and is checked before payment.
Product description
ANAND AI · DeepSeek API, no subscription
DeepSeek API key with a $1 balance. Works with any OpenAI-compatible client: Cursor, Cline, Continue, LobeChat, Open WebUI, Hermes and other agents, or your own code.
STEP 1. TAKE IT FROM YOUR DELIVERY
Together with the key you get three lines: Base URL, API Key and model names. The key looks like sk-... It is personal: do not share it. If it leaks, message me in the marketplace chat and I will revoke it and issue a new one.
STEP 2. PASTE ADDRESS AND KEY INTO YOUR CLIENT
Base URL — from your delivery, API Key — same place.
Model: deepseek-flash (fast, cheap) or deepseek-v4-pro (hard tasks)
- Cursor: Settings - Models - OpenAI API Key and Override OpenAI Base URL, then add the model to the list.
- Cline / Continue in VS Code: Provider - OpenAI Compatible, same address, key and Model ID.
- LobeChat / Open WebUI: Settings - Language model - OpenAI, API address, key, model.
- Your own code: OpenAI(base_url="address from delivery", api_key="sk-...") — same request format as OpenAI.
FOR AGENTS: HERMES, CODEX, AUTOGEN, LANGCHAIN
Tested on Hermes Agent: plain answers and tool calls both work (tools / function calling are supported).
Quick start via environment variables:
CUSTOM_BASE_URL=address_from_delivery
CUSTOM_API_KEY=sk-your_key
hermes -m deepseek-flash --provider custom
Permanent setup: put the same two lines into ~/.hermes/.env, then just run hermes.
Via config.yaml:
model:
default: deepseek-flash
provider: custom
base_url: address from your delivery
No config editing needed either: run hermes model and pick custom / OpenAI-compatible, then paste address and key in the menu.
STEP 3. CHECK THAT IT WORKS
GET /models with the header Authorization: Bearer your_key must return a list of two models. This call is free and does not spend your balance.
HOW TO CHECK YOUR BALANCE
Easiest — the balance page: its link arrives in your delivery together with the key. Paste the key and you get the remaining balance, spend, a daily chart and model prices on one screen. The key stays in your browser: the page does not store it and sends it nowhere.
For scripts: GET the same address you received, path /key/info, with the same Authorization: Bearer your_key header. The answer contains max_budget (what the key was loaded with) and spend (already used). Remaining = max_budget minus spend. This call is free and spends nothing.
Once the limit is used up the API starts returning 400 budget exceeded — the balance is gone: message me in the marketplace chat, I will sell the next denomination and you can keep the old key.
IF IT DOES NOT WORK
401 Unauthorized — the key was copied partially or without sk-. Copy the whole line, no spaces.
404 or model not found — the Base URL is missing /v1, or the model name is different: only deepseek-flash and deepseek-v4-pro.
400 about context length — shorten the history; in chat clients enable context trimming.
Empty answer with reasoning_content — the model reasoned and hit max_tokens, raise the limit.
429 — too many requests, wait and retry.
Still stuck — message me in the marketplace chat with the error text and I will replace the key.
PROMPT CACHE IS ON
Send the same context again — system prompt, project files, a long instruction — and DeepSeek serves it from cache, where input costs 10x less. In a live check 94% of the input was billed at the cache price on the second identical request. Nothing to configure, the cache works by itself. It lives for minutes to hours: after a long break the first request warms the context up again.
SPECS
Context: up to 1,000,000 input tokens, up to 393,216 output tokens.
Prices per 1M tokens: deepseek-v4-pro input $1.32 / output $3.96; deepseek-flash input $0.30 / output $1.20. Prompt cache is 30x cheaper (pro) and 50x (flash). Off-peak hours are half price.
Peak hours: 01:00-04:00 and 06:00-10:00 UTC on weekdays.
A $1 balance, no subscription and no commitment: enough to try it out, and larger denominations cost less per token. The balance is spent only on your requests and never expires. Questions — in the marketplace chat, I answer within the day.