Price per Million Tokens: Budgeting AI Costs in Practice

By Darus S Alfatah · 2025-09-20 · Edukasi · 8 menit

Price per Million Tokens: Budgeting AI Costs in Practice

Author note: Always read input and output prices separately; the gap is often two to five times.

Focus of this article

This article compares per-million-token rates across providers so you can convert prices into the same unit.

Read also: How to Calculate AI Cost per Project (A Real Case Study)

Why the Unit Matters

Prices are quoted per million tokens because a single request costs a fraction of a cent. That unit helps comparison but not planning. Translate it into a monthly figure for your own traffic and the numbers start to drive decisions.

Two applications can pay the same price per token and have very different bills, because one sends long contexts and the other sends short questions.

The Formula

Monthly cost equals requests multiplied by average tokens per request, multiplied by price per token. Separate input and output if your provider prices them differently, which most do.

Add a margin for retries and experiments. Twenty percent is a reasonable starting point for a new feature, because first-month usage is rarely stable.

Example One: FAQ Chatbot

Two thousand conversations with eight hundred tokens each is about 1.6 million tokens per month. Here the cheapest adequate model is almost always the right choice, and the total stays small regardless of provider.

Cache repeated questions and the cost falls further, often by half.

Example Two: Content Assistant

Five hundred requests with long context at three thousand tokens each is around 1.5 million tokens. Quality expectations are higher, so a mid-tier model is justified, but retrieving only relevant references keeps the input from growing.

Example Three: Classification

Ten thousand short requests at two hundred tokens each is two million tokens. Volume is high but each request is trivial, so a small model handles it well and the cost stays predictable.

Batch these when latency allows; batching usually reduces price further.

What Raises Bills Unexpectedly

Retries, long conversation history, duplicated system prompts, and experiments run outside the tracked pipeline. Each is small alone and significant in aggregate.

Set a daily budget alert and review the three most expensive features monthly. That routine catches most surprises before they compound.

Frequently Asked Questions

Is input or output more expensive? Usually output, often two to four times input.

How many tokens is an Indonesian sentence? Roughly one and a half to two per word.

Do cached tokens cost less? Yes, often substantially.

Should I buy a bigger package? Only when usage is stable; otherwise pay as you go.

How do I compare providers fairly? Compute cost per successful task, not per token.

Conclusion

Convert prices into monthly figures for your own traffic, add margin for retries, and track the top features. That is enough to budget without surprises.

Building a Simple Budget Sheet

Keep a sheet with one row per feature: model, average input tokens, average output tokens, monthly requests, and total cost. Update it monthly from real logs rather than estimates. Within two months you can predict next month's bill within a small margin.

Share the sheet with the team. Cost awareness spreads fastest when people can see their own feature in the list.

When to Negotiate or Switch

If a single feature dominates spend, it deserves an optimisation pass before any provider negotiation. Trimming context and caching usually saves more than haggling, and it improves latency as a side effect.

What the number means in practice: One million tokens is roughly hundreds of thousands of words. For scale, a thick novel is on the order of hundreds of thousands of tokens.

Frequently Asked Questions

What is the core topic of this article?

This article explains the fundamentals of AI APIs and gateways in plain language for developers, with practical examples.

Do I need a technical background?

No. The material builds up from the basics, so beginners can follow while experienced developers can jump to the relevant section.

How do I get started with an AI API after reading this?

Create a BankAI account, get an API key, then change your Base URL to BankAI’s OpenAI-compatible endpoint.

Sources & References

About the author

Darus S Alfatah — Founder & Writer

Darus S Alfatah is the founder of BankAI. He builds and operates the LLM API gateway that serves Indonesian developers, and writes about LLM integration, gateway architecture, and using AI practically and cost-effectively. Every guide is tested in a real environment — including code, configuration, and token calculations — before it is published, and updated when provider documentation or pricing changes.

Editorial policy

Related articles

🚀 Ready to Use the All-in-One API Key?

Sign up, top up with QRIS, and start building with ChatGPT, Claude, DeepSeek and every other model through one key.

🔑 SIGN UP NOW · Pricing