EN ▾
Uncensored Chatbot APIDirect API access to one uncensored LLMGet API key

Uncensored Chatbot APISafety

Chatbot on Website Cost: Real Numbers Per Chat

The chatbot on website cost you pay with a pay-per-token API is about half a cent per conversation. Below are the numbers, how they scale with traffic, and what makes a bill grow.

Updated

Browser window with a chat bubble feeding coins into a jar, showing per-chat token cost on a website chatbot

Key points

  • A typical 10-turn chat costs about $0.0045, so 1,000 chats cost about $4.51.
  • Pricing is $0.15 per 1M input tokens and $0.50 per 1M output tokens, with no subscription.
  • Resent chat history is the main cost driver, so trim it and keep the system prompt short.
  • Top-ups start at $20, credit never expires, and a $0.50 trial lets you measure your own usage first.

The short version

If you run a chatbot on your website using a prepaid, pay-per-token API, you do not pay a monthly fee. You pay for tokens: $0.15 per 1M input tokens and $0.50 per 1M output tokens. For a normal support-style or character chat of about ten turns, the total is roughly half a cent. A site that handles 4,000 such chats in a month spends about $18.05.

Your real bill depends on three things: how long each conversation runs, how much history you send back with every message, and how long the replies are. The rest of this article shows how to compute each one so you can budget before you build.

How tokens turn into dollars

A token is a chunk of text, roughly three quarters of an English word. Every request has an input side (the system prompt, the earlier messages and the new question) and an output side (the reply). Input costs $0.15 per million tokens and output costs $0.50 per million, so a token you receive costs a bit over three times what a token you send does.

The formula is simple:

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

A single question with 800 input tokens and a 250-token reply costs 800 x 0.15 / 1,000,000 = $0.00012 for input and 250 x 0.50 / 1,000,000 = $0.000125 for output. Together that is $0.000245. At that size you could send about 4,000 requests for one dollar. Failed or refused requests are not charged, so errors do not inflate the figure.

A realistic 10-turn conversation

Chat is where people underestimate cost, because each new message resends the earlier ones. Take a concrete setup:

  • System prompt: 500 tokens, sent on every request.
  • Each user message: 100 tokens.
  • Each reply: 250 tokens.
  • Conversation length: 10 turns.

On turn 1 the input is 500 + 100 = 600 tokens. Each later turn adds 350 tokens of history (one user message plus one reply), so turn k has an input of 500 + (k - 1) x 350 + 100. Summing turns 1 to 10 gives 10 x 600 + 350 x (0 + 1 + ... + 9) = 6,000 + 350 x 45 = 21,750 input tokens. Output is 10 x 250 = 2,500 tokens.

  • Input: 21,750 x 0.15 / 1,000,000 = $0.0032625.
  • Output: 2,500 x 0.50 / 1,000,000 = $0.00125.
  • Whole conversation: $0.0045125, call it $0.0045.

Notice that the input side is more than two thirds of the cost even though input is the cheaper token. That is the effect of resending history, and it grows faster than the conversation does.

Monthly estimates by traffic

Using the $0.0045125 per conversation figure above:

Chats per monthApproximate cost
1,000$4.51
4,000$18.05
20,000$90.25
100,000$451.25

These are model costs only. Your web hosting, domain and any moderation tools are separate, and I have no figures for them, so check your own providers. The model side is usually the smaller line for a small site and the larger one for a busy site.

Longer chats cost more than proportionally. Do the same math for 20 turns with the same message sizes and the input side climbs to about 78,500 tokens, because the sum of history grows with the square of the turn count. That is why history management, covered below, is the best lever you have.

Five ways to keep the bill small

  1. Trim history. Send only the last 6 to 10 messages, plus a short running summary if the chat needs memory. Dropping old turns cuts the squared growth.
  2. Keep the system prompt tight. A 500-token prompt over 10 turns is 5,000 input tokens, or $0.00075 per conversation. A 2,000-token prompt is four times that. Move rarely needed rules into a short note that you include only when relevant.
  3. Cap reply length. Ask for short answers in the prompt, and set max_tokens when you want a hard ceiling. It is optional, since the completion can otherwise use the rest of the context window.
  4. Limit conversation length per visitor. A soft cap of 30 messages per session protects you from one visitor running up a large history.
  5. Rate-limit per visitor in your own backend. The API caps a full key at 8 concurrent requests and 300 requests per minute, but it cannot tell your visitors apart. Your server can.

The web chat tutorial shows where the history trimming goes in a working page.

Three example budgets

The 10-turn chat is one shape of traffic. Here are two others, with the math, so you can see how much the shape matters.

An FAQ bot with single questions

Each request carries 1,000 input tokens (a short instruction plus a snippet from your help page) and returns 150 tokens. Input costs 1,000 x 0.15 / 1,000,000 = $0.00015, output costs 150 x 0.50 / 1,000,000 = $0.000075, and the total is $0.000225. Ten thousand questions cost $2.25. There is no history to resend, which is why this shape is the cheapest.

A companion-style chat with a trimmed window

A long-running character chat of 30 turns, with the last ten messages kept, might average 3,000 input tokens and 300 output tokens per turn. Over 30 turns that is 90,000 input and 9,000 output tokens: $0.0135 plus $0.0045, so $0.018 per session. One thousand such sessions cost $18.00. The window cap keeps the input flat per turn instead of growing, so a 30-turn session costs about four times a 10-turn session instead of growing with the square of its length.

What this tells you

Cost per chat moves by a factor of twenty between the FAQ bot and the companion app, while the price per token never changes. Estimate by shape of use, not by the number of visitors alone.

Paying: top-ups, bonuses and the trial

You fund the account with prepaid credit. Any whole amount from $20 to $500 works, and larger top-ups add a bonus: +5% from $50, +10% from $100, +15% from $200. A $100 top-up becomes $110 of credit, which covers about 24,400 of the 10-turn chats above (110 divided by 0.0045125). The smallest top-up of $20 covers about 4,430 chats.

Payment is by crypto: USDT on TRC20 or USDC on Base. Credit never expires and there is no subscription, so a quiet month costs nothing beyond what you used. The response header X-Balance-USD returns your balance on every call, which makes a low-balance alert easy to add.

Before you top up, claim the trial. New users can get $0.50 of trial credit with one click on the API key page after signing up, one per person, valid for 7 days, no payment needed. At the example rate, $0.50 covers about 110 full conversations, enough to test with real prompts. A trial key allows 2 concurrent requests and 60 per minute. The free trial FAQ covers the rules. Full pricing is on the pricing page.

Measure your own numbers

Estimates are only a starting point. The API returns usage data (with streaming, usage arrives at the end), so you can log real input and output counts per chat and compute true cost. This short function does the math from a response:

from openai import OpenAI

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

def cost_usd(usage):
    return usage.prompt_tokens * 0.15 / 1_000_000 + usage.completion_tokens * 0.50 / 1_000_000

resp = client.chat.completions.create(
    model="uncensored",
    messages=[{"role": "user", "content": "Give me three opening lines for a support greeting."}],
)
print(resp.usage.prompt_tokens, resp.usage.completion_tokens, round(cost_usd(resp.usage), 6))

Run a week of real traffic, average the cost per chat, and replace my numbers with yours. Keep prompts and logs in line with your privacy page, and if your bot handles adult topics, read the notes on responsible usage and safeguards.

Questions and answers

How much does a chatbot on a small website cost per month?

At about $0.0045 per 10-turn conversation, 1,000 chats a month cost about $4.51 and 4,000 cost about $18.05. Your own number depends on chat length and prompt size.

Is there a monthly fee?

No. You pay per token from prepaid credit. The minimum top-up is $20 and the credit never expires.

What is the biggest cost driver?

Resent history. Each new message includes earlier ones as input, so long chats cost more than proportionally. Trimming old turns helps most.

Can I test before paying?

Yes. New users can claim $0.50 of trial credit with one click on the API key page, valid for 7 days. It covers roughly 110 chats of the size used here.

Your key is one form away

Create an account, copy the key, change the base URL. That is the whole setup.

Get API keyRead the docs