AI API Key Pricing: What One Key Really Costs

A key is free, tokens are not

Many people search for ai api key pricing expecting a monthly fee per key. That is not how usage-based LLM APIs work. A key is only a credential that identifies your account. You can create one without paying, and it costs nothing to hold.

The bill comes from tokens. A token is a small chunk of text, roughly three quarters of an English word. Every request has input tokens (your prompt, the conversation history, any system message) and output tokens (the reply). Here the rates are:

• Input: $0.15 per 1M tokens
• Output: $0.50 per 1M tokens

There is no subscription and no minimum monthly spend. If you send nothing, you pay nothing. The balance sits in your account until you use it, and credit never expires.

How to calculate the cost of one request

The formula is short:

cost = input_tokens x 0.15 / 1,000,000 + output_tokens x 0.50 / 1,000,000

Take a request with 1,000 input tokens and 500 output tokens. Input costs 1,000 x 0.15 / 1M = $0.00015. Output costs 500 x 0.50 / 1M = $0.00025. The request costs $0.0004 in total, which is four hundredths of a cent. Put differently, $1 pays for 2,500 requests of that size.

Output tokens cost more than input tokens, so a long answer to a short question is more expensive than a short answer to a long document. When you want to cut cost, ask for shorter replies first, then shorten the history you send each turn. Every response reports its token usage, and streamed responses include usage at the end, so you can log real numbers instead of estimating.

Monthly estimates for three kinds of use

These are worked examples, not promises. Replace the request counts and token sizes with your own. All use a 30-day month.

Profile	Requests per day	Tokens per request (in / out)	Monthly cost
Hobby project or personal tool	200	1,000 / 500	$2.40
Community bot	3,000	2,000 / 400	$45.00
Production product	50,000	3,000 / 600	$1,125.00

The arithmetic

• Hobby: 200 x 30 = 6,000 requests. Input 6M tokens x $0.15 = $0.90. Output 3M tokens x $0.50 = $1.50. Total $2.40.
• Bot: 3,000 x 30 = 90,000 requests. Input 180M tokens = $27.00. Output 36M tokens = $18.00. Total $45.00.
• Production: 50,000 x 30 = 1.5M requests. Input 4,500M tokens = $675. Output 900M tokens = $450. Total $1,125.

The bot case is dominated by input, because each message carries a long prompt and chat history. The hobby case is dominated by output. Knowing which side dominates tells you where to optimise.

Adding credit: amounts and bonuses

You top up with any whole amount from $20 to $500. Larger top-ups carry a bonus:

Top-up	Bonus	Credit you get
$20 to $49	none	The amount you pay
$50	+5%	$52.50
$100	+10%	$110.00
$200	+15%	$230.00

You can pay by card (Visa, Mastercard, Apple Pay, Google Pay) or with crypto (USDT on TRC20, USDC on Base). Because credit never expires, a bigger top-up with a bonus is a straightforward saving if you know you will use it, and no loss if you use it slowly. The pricing page lists the current amounts.

Try it for $0.50 first

New users can claim $0.50 of trial credit with one click on the API key page after sign-up. It needs no card, it is limited to one per person, and it is valid for 7 days. At the hobby request size above, $0.50 covers 1,250 requests, which is enough to test a prompt, a bot command or a small script properly.

A trial key has lower limits (2 concurrent requests and 60 per minute) than a full key (8 and 300). If you are still looking at the range of free options, the article on getting a free LLM API key compares what free actually buys you. When you are ready, create your key here.

Track real cost from the usage field

Estimates are useful for planning, but your own logs are better. Each non-streamed response includes a usage object with prompt and completion token counts. This helper turns them into dollars:

Copyfrom openai import OpenAI

client = OpenAI(base_url="https://api.getaiapikey.com/v1", api_key="YOUR_KEY")

IN_RATE = 0.15 / 1_000_000   # dollars per input token
OUT_RATE = 0.50 / 1_000_000  # dollars per output token

r = client.chat.completions.create(
    model="uncensored",
    messages=[{"role": "user", "content": "Write a two-line product blurb for a mug."}],
)
u = r.usage
cost = u.prompt_tokens * IN_RATE + u.completion_tokens * OUT_RATE
print(u.prompt_tokens, u.completion_tokens, f"${cost:.6f}")

Run it over a day of real traffic and multiply. If the number is far from your table estimate, your average prompt is longer or shorter than you assumed, and that is the first thing to correct.

Why pay per token instead of a flat plan

A flat monthly plan suits steady heavy use. Usage-based billing suits everything else: a bot that is busy on weekends, a prototype that sits idle for three weeks, a script you run once a month. With prepaid credit that never expires, a quiet month costs nothing extra, and you never pay for capacity you did not use.

What is and is not included in the price

The per-token rate covers the one model available on this API (model ID uncensored), a single open-weight model run by an independent provider. There is no charge for the key, for the dashboard, or for switching between SDKs. Requests that fail or are refused are not charged, so retries after errors do not add to your bill.

Some things are not offered at all, so they are not in the price: image, audio and video generation, embeddings and fine-tuning. The service is text only. Context is 100,000 tokens for prompt plus completion, and a very long prompt costs more simply because it contains more input tokens. At the full window, one request with 100,000 input tokens would cost $0.015 for the input alone.

If you are comparing against another provider, check their current pricing on their own site, and compare on the same basis: input rate, output rate, minimum spend and whether credit expires. A cheap headline rate with an expiring balance is not the same deal as a slightly higher rate with credit that stays.

Ways to keep the bill small

• Trim history. Chat apps resend the whole conversation each turn. Keep the last 10 to 20 messages, or summarise older ones.
• Ask for brevity. A system line such as "answer in at most three sentences" cuts output tokens, the expensive side.
• Set max_tokens when you know the maximum useful length. It is optional here, but it protects you from a runaway reply.
• Log usage per feature. One chatty feature is often most of the bill.
• Watch the balance. The X-Balance-USD header on every response lets you alert yourself before the credit runs out.

The overview of the artificial intelligence API shows how these pieces fit together, and the documentation has the request format.

More guides

• AI API Key Tester: Check a Key in 30 Seconds
• AI API Checklist