Guide · DeepSeek API
How to use the DeepSeek API in Pakistan (pay in rupees)
Updated · Habi Code
DeepSeek's models are strong at chat and coding and cheap per token, but most AI providers bill in US dollars and expect an international card. This guide shows how to use DeepSeek V4 from Pakistan and pay in rupees through Habi Code, with code you can copy.
Habi Code is an independent service based in Pakistan. It is not part of DeepSeek. It gives you one OpenAI-compatible API key, a rupee balance you top up with JazzCash, Easypaisa or a bank transfer, and a price per million tokens in rupees.
Which DeepSeek models you can use
Three DeepSeek V4 models are available today, plus GLM 5.3 Flash and MiniMax M2.7 on the same key. The table below is the current price list:
| Model | Model id | Price per 1M tokens | Longest reply | Rs 500 buys about |
|---|---|---|---|---|
| DeepSeek V4 Flash | deepseek-v4-flash | Rs 10 (prompt and reply) | 16K tokens (about 12,000 words) | 50M tokens |
| DeepSeek V4 Fast | deepseek-v4-fast | Rs 50 (prompt and reply) | 393K tokens (about 295,000 words) | 10M tokens |
| GLM 5.3 Flash | glm-5.3-flash | Rs 10 (prompt and reply) | 16K tokens (about 12,000 words) | 50M tokens |
| MiniMax M2.7 | minimax-m2.7 | Rs 10 (prompt and reply) | 16K tokens (about 12,000 words) | 50M tokens |
| DeepSeek V4 Uncensored | deepseek-v4-flash-long | Rs 100 (prompt and reply) | 393K tokens (about 295,000 words) | 5M tokens |
Coming soon: GPT-6.1 Terra, GPT-6.1 Sol, GPT-6 Astra.
Deal: Rs 1,000 = 100M tokens on DeepSeek V4 Flash, GLM 5.3 Flash and MiniMax M2.7 (prompt, reply and cache all count the same; one-time top-up, never expires).
Packs (one-time top-ups; the balance works on every model and never expires): Rs 500 = about 50M DeepSeek V4 Flash, 10M DeepSeek V4 Fast or 5M DeepSeek V4 Uncensored tokens; Rs 1,000 = about 100M DeepSeek V4 Flash, 20M DeepSeek V4 Fast or 10M DeepSeek V4 Uncensored tokens; Rs 2,500 = about 250M DeepSeek V4 Flash, 50M DeepSeek V4 Fast or 25M DeepSeek V4 Uncensored tokens.
Free trial: new accounts get Rs 10 credit (a limited number each day, one per internet connection).
Pay with JazzCash, Easypaisa or bank transfer, then send the payment screenshot on WhatsApp; the balance is added after the payment is checked. Base URL for OpenAI-compatible code: https://p01--caddy--pjd5xh74ttqd.code.run/v1. Prices as of 2026-11-19.
- DeepSeek V4 Flash (
deepseek-v4-flash): the everyday choice for chat, coding help and bots. Replies can be up to 16K tokens, about 12,000 words. - DeepSeek V4 Fast (
deepseek-v4-fast): for very long answers, such as full documents or large code files. Replies can be up to 393K tokens. - DeepSeek V4 Uncensored (
deepseek-v4-flash-long): no topic filters, with the same 393K reply limit. Uncensored does not mean anything goes: illegal use is not allowed under the terms.
Each price is for the prompt and the reply together: you pay the same rate for the tokens you send and the tokens you get back. One token is about three quarters of an English word.
Step 1: create an account
Open your account page and press Get started. There is no email and no password: you get a secret code, or you can use your WhatsApp number and a PIN. Keep the code safe, because it is your account. New accounts get a Rs 10 trial credit (a limited number each day, one per internet connection), which is about 1M tokens on DeepSeek V4 Flash, enough to test your code.
Step 2: top up in rupees
The smallest top-up is Rs 500. You pay by JazzCash, Easypaisa or bank transfer, send the payment screenshot with your account ID on WhatsApp, and the balance is added after the payment is checked. The full steps are in the payment guide. Your balance stays in rupees, never expires and works on every model.
Step 3: make an API key
In the dashboard, create an API key and copy it. You can make more than one key (for example one per project) and delete a key you no longer trust.
Step 4: call DeepSeek from your code
Because the API is OpenAI-compatible, you use the official OpenAI library and change only two things: the base URL and the key. The base URL is:
https://p01--caddy--pjd5xh74ttqd.code.run/v1
Python
Install the library with pip install openai, then:
from openai import OpenAI
client = OpenAI(
base_url="https://p01--caddy--pjd5xh74ttqd.code.run/v1",
api_key="sk-your-habicode-key",
)
reply = client.chat.completions.create(
model="deepseek-v4-flash",
messages=[{"role": "user", "content": "Explain recursion in two lines."}],
max_tokens=500,
)
print(reply.choices[0].message.content)
JavaScript (Node.js)
Install with npm install openai. This example uses DeepSeek V4 Fast for a long reply:
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://p01--caddy--pjd5xh74ttqd.code.run/v1",
apiKey: "sk-your-habicode-key",
});
const reply = await client.chat.completions.create({
model: "deepseek-v4-fast",
messages: [{ role: "user", content: "Write a long README for my project." }],
});
console.log(reply.choices[0].message.content);
curl
curl https://p01--caddy--pjd5xh74ttqd.code.run/v1/chat/completions \
-H "Authorization: Bearer sk-your-habicode-key" \
-H "Content-Type: application/json" \
-d '{"model": "deepseek-v4-flash", "messages": [{"role": "user", "content": "Salam!"}]}'
To switch models, change the model value to another id from the table. Everything else stays the same.
Using it in coding tools
Tools that accept an "OpenAI-compatible" provider, such as Cline and Continue, work the same way: choose the OpenAI-compatible option, paste the base URL and your key, and type a model id such as deepseek-v4-flash. If you would rather not set anything up in an editor, the free Habi Code desktop app for Windows is a coding agent that runs on the same balance and key.
What it costs in practice
At Rs 10 per million tokens on DeepSeek V4 Flash:
- A question and answer that use 2,000 tokens in total cost Rs 0.02.
- One full 16K-token reply costs Rs 0.16.
- Rs 500 buys about 50 million tokens.
On DeepSeek V4 Fast (Rs 50 per million), a full 393K-token reply costs about Rs 19.65; on DeepSeek V4 Uncensored (Rs 100 per million) it costs about Rs 39.30. Every request in your dashboard shows its tokens and its exact cost, so you can check the maths yourself.
Tips to keep the bill small
- Set
max_tokenswhen you do not need a long answer. It caps the reply length. - Remember the prompt counts too. In a chat, the whole history is sent again with every message, so long conversations cost more per turn. Start a new conversation when the topic changes.
- Pick the cheapest model that does the job. Use Flash by default and switch to Fast only when you need replies longer than 16K tokens.
- Test with a small amount first, then top up more once your code works.
If something goes wrong
If you get an authentication error, check that the key is copied in full and that the base URL ends in /v1. If the model is not found, check the model id: it must match the table exactly, in lower case. If requests fail because your balance has run out, top up again; the new amount is added to what you already have. A request that fails with an error is normally not charged (see the refund policy).