How to Pay for AI API Usage in INR
AI APIs charge per token, and on Velona every price is shown in rupees. As of 4 October 2026, 1M input tokens cost ₹2.90 on DeepSeek V4 Flash, ₹15.54 on GPT-4o mini, ₹31.08 on Gemini 2.5 Flash and ₹207.22 on Claude Sonnet 5. A typical chat reply costs anywhere from under 1 paisa to about 73 paise depending on the model.
This guide explains what you're paying for, shows real rupee costs, and gives you a simple way to estimate a monthly bill.
What is a token?
A token is a small chunk of text the model reads or writes. In English, one token is roughly three quarters of a word, so 1,000 tokens is about 750 words. Hindi and other Indian languages often use noticeably more tokens for the same meaning, so non-English apps can cost more per message.
You pay for two kinds of tokens:
- Input tokens: everything you send, including your instructions, the chat history and any documents.
- Output tokens: everything the model writes back. These usually cost several times more.
Token prices in rupees (₹ per 1M tokens)
| Model | Input ₹ / 1M | Output ₹ / 1M |
|---|---|---|
| DeepSeek V4 Flash | ₹2.90 | ₹5.80 |
| GPT-OSS 120B | ₹3.83 | ₹17.61 |
| GPT-4o mini | ₹15.54 | ₹62.17 |
| DeepSeek V4 Pro | ₹21.63 | ₹43.27 |
| Gemini 2.5 Flash | ₹31.08 | ₹259.02 |
| Claude Haiku 4.5 | ₹103.61 | ₹518.05 |
| GPT-5 | ₹129.51 | ₹1,036 |
| Claude Sonnet 5 | ₹207.22 | ₹1,036 |
| Claude Opus 4.8 | ₹518.05 | ₹2,590 |
Prices follow the live exchange rate and include Velona's 7% platform fee and 0.5% forex buffer. Every model's current price is on the model catalogue, and the formula is explained in our pricing explainer.
How to calculate the cost of one request
Cost in ₹ = (input tokens × input price ÷ 1,000,000)
+ (output tokens × output price ÷ 1,000,000)Example: a request to Claude Sonnet 5 with 6,000 input tokens and 600 output tokens:
6,000 × 207.22 ÷ 1,000,000 = ₹1.24
600 × 1,036 ÷ 1,000,000 = ₹0.62
Total = ₹1.86The same request on DeepSeek V4 Flash costs about ₹0.02.
What common requests cost
| Request | Tokens (in / out) | DeepSeek V4 Flash | GPT-4o mini | Claude Sonnet 5 |
|---|---|---|---|---|
| Chat reply | 1,000 / 500 | under ₹0.01 | about ₹0.05 | about ₹0.73 |
| Summarise 10 pages | 6,000 / 600 | about ₹0.02 | about ₹0.13 | about ₹1.86 |
| Coding agent step | 30,000 / 1,000 | about ₹0.09 | about ₹0.53 | about ₹7.25 |
Estimate your monthly bill in rupees
Use four numbers: users, messages per user per day, days, and average tokens per message.
Example: 100 users each send 10 messages a day for 30 days, and each message uses about 1,000 input tokens and 500 output tokens.
Requests: 100 × 10 × 30 = 30,000
Input tokens: 30,000 × 1,000 = 30M
Output tokens: 30,000 × 500 = 15M| Model | Monthly cost |
|---|---|
| DeepSeek V4 Flash | about ₹174 |
| GPT-4o mini | about ₹1,400 |
| Claude Sonnet 5 | about ₹21,760 |
The model you choose changes the bill by more than 100 times for the same traffic, which is why most apps run on a cheap model and use a premium one only where it matters.
Track cost in your own code
Every response tells you the tokens used. You can turn that into rupees:
PRICES = { # ₹ per 1M tokens: (input, output)
"deepseek/deepseek-v4-flash": (2.90, 5.80),
"openai/gpt-4o-mini": (15.54, 62.17),
"anthropic/claude-sonnet-5": (207.22, 1036),
}
def cost_inr(model, usage):
inp, out = PRICES[model]
return (usage.prompt_tokens * inp + usage.completion_tokens * out) / 1_000_000Keep prices in one place like this, because they change with the exchange rate. Your usage page always shows the exact rupee cost of every request.
Five ways to pay less per request
- Start on a cheap model and upgrade only the steps that need it.
- Cap output with
max_tokens. Output is the expensive side. - Send less history. Chat apps resend old messages every time, so keep only the recent ones.
- Don't paste whole documents when a relevant section will do.
- Watch for loops. A bug that repeats requests is the most common cause of surprise bills. A budget on each API key stops it.
Frequently asked questions
How much is 1M tokens in rupees?
It depends on the model: from ₹2.90 (DeepSeek V4 Flash input) to over ₹500 (Claude Opus 4.8 input) per 1M tokens on Velona today.
How many words is 1M tokens?
About 750,000 English words, roughly ten average novels.
Why is output more expensive than input?
Generating text takes more computation than reading it, so providers price output tokens higher.
Can I pay for tokens in rupees?
Yes. Velona prices every model in rupees and you top up with UPI from ₹10. See how to pay for AI APIs with UPI.
Create your Velona account and see the rupee cost of every request from your first call.