Claude, GPT, Gemini, DeepSeek API prices in rupiah
How to read AI API rates per 1 million tokens and convert them to rupiah, with worked examples for one request and one coding session.
AI model API prices are almost always written in USD per 1 million tokens. For developers in Indonesia, that number is hard to picture: how many rupiah for one question, or for an afternoon with a coding agent? This post explains how to read rates per 1 million tokens and convert them to rupiah, with worked examples you can repeat yourself.
Where to see prices per model
The rates for every model on Cent AI, including models from the Claude, GPT, Gemini and DeepSeek families, are on the pricing page. That page shows the Cent AI rate, the model maker's official price and the actual discount. The model list can be filtered by family on the models page, and the same rates are returned by GET /v1/models.
We deliberately do not copy per-model rates into this post. Prices can go down or up, and every change is announced in the changelog with the old rate, the new rate and when it takes effect. The pricing page always shows the version in force.
The four rates to read
Every model has four rates per 1 million tokens:
| Rate | What it charges for |
|---|---|
| Input | Prompt tokens the model reads fresh |
| Cache read | Prompt tokens served from the model's cache |
| Cache write | Prompt tokens stored in the cache |
| Output | Tokens the model writes, reasoning tokens included |
On Cent AI, cache read is 10% of the model's input rate, and cache write is billed at the input rate. Output is usually the most expensive rate per token, but there are far fewer output tokens than input tokens.
Converting to rupiah
The formula is simple:
rupiah per 1M tokens = USD rate per 1M tokens × exchange rate
For the examples below we use an illustrative rate of about Rp16,500 per $1. It is only an estimate. When you top up with QRIS, Cent AI uses the live rate with no spread added, locks it when the order is created, and shows it before you pay.
Example: one request
The rates in this section are examples to show the arithmetic, not the price of any particular model. Say a model charges $0.10 for input, $0.01 for cache read and $0.50 for output per 1 million tokens. In rupiah, at the illustrative rate, that is about Rp1,650 for input, Rp165 for cache read and Rp8,250 for output, per 1 million tokens.
Now take one request with 50,000 prompt tokens, 40,000 of them read from the cache, and 2,000 output tokens:
| Part | Tokens | Cost in USD |
|---|---|---|
| New input | 10,000 | $0.001000 |
| Cache read | 40,000 | $0.000400 |
| Output | 2,000 | $0.001000 |
| Total | $0.002400 |
In rupiah, $0.0024 is about Rp40. Cost is calculated in billionths of a dollar and rounded up once per request. A successful request has a minimum charge of $0.00002, and a request that fails before the model accepts it is not charged.
Example: one coding session
A coding agent resends the conversation on every turn, so a pattern like the one above repeats dozens of times. At the same example rates, 100 such turns cost $0.24, about Rp4,000. If the cache never hits, the 40,000 prompt tokens that were cache reads are billed as full input, and the cost per turn rises to $0.0060, about Rp99. That is why the cache matters so much for coding agent costs.
Real numbers depend on the model, the context length and how often the cache hits. Cent AI does not promise a cache hit rate, and cache tokens are billed only when the model reports them.
Try it with real rates
The calculator below uses the same price list as the pricing page. Pick a model, change the token counts and the cache share, then multiply the result by the exchange rate for a rupiah estimate.
Cost per request
Estimate
deepseek-v4.1-flashAn estimate at price version 4. Your bill comes from the usage the provider reports for each request.
Save $0.0195Comparing models fairly
When you compare models, do not look at the input rate alone. A few things shape the real cost:
- Cache share. A model with a higher input rate can cost less per session if its cache hits often.
- Output and reasoning length. Reasoning tokens are billed at the output rate. A model that thinks a lot produces more output tokens.
- Minimum per request. For very short requests, the per-request minimum can be larger than the token cost.
Paying in rupiah
A Cent AI balance is prepaid USD, but you can fund it in rupiah with QRIS from $1, or with USDT/USDC on Solana (the network fee charged by your wallet is yours). There are also one-time budget packages of $1.99, $9.99 and $99 that do not expire while the account is active; each package size covers a set of models listed on the pricing page.
Once requests run, every row on the dashboard's Requests page shows tokens per rate and the cost, and every response carries usage.cost. The full formula is in the pricing docs.
Sources
- cent-ai.dev/docs/concepts/pricing (opens in a new tab) · accessed 2026-11-14
- cent-ai.dev/docs/concepts/top-up (opens in a new tab) · accessed 2026-11-14
- cent-ai.dev/pricing (opens in a new tab) · accessed 2026-11-14
Related posts
- Anthropic prompt caching pricing, worked outpricing · 4 minHow Anthropic's cache write and cache read are priced, which usage fields carry them, and a worked usage.cost example for two requests that share one cached prompt.
- Four rates per model, and what each one chargespricing · 2 minInput, cache read, cache write and output each have their own price. How one request is billed, step by step, and why the result is rounded up once.
- Card declined when paying for an AI API? What to checkpayment · 3 minCommon things to check when a debit or credit card is declined while buying AI API credits, and a prepaid alternative with QRIS or USDT/USDC.