EN ▾
Get API key

LLM Inference APIGuide

LLM API Cost per Month: Four Realistic Scenarios

Four monthly bills worked out line by line at $0.15 per 1M input and $0.50 per 1M output tokens, plus a script to price your own traffic.

Updated:

Four rising glass bars filled with glowing cubes, one per monthly cost scenario
Isometric illustration of data packets flowing from a terminal into a cloud gateway, representing an inference API call

Key points

  1. Point any OpenAI SDK at https://api.llminferenceapi.com/v1, use the model ID uncensored and send your key as a Bearer token.
  2. Two endpoints are available: POST /v1/chat/completions and GET /v1/models.
  3. Streaming, tool calling and JSON mode work through the standard request fields.
  4. A full key allows 8 concurrent requests and 300 per minute; a 429 response tells you which limit you hit.

The four scenarios at a glance

If you are trying to work out llm api cost per month, a single average is no help, because bills depend on how many requests you send and how long they are. Below are four realistic setups priced at $0.15 per 1M input tokens and $0.50 per 1M output tokens. Each one shows its assumptions so you can swap in your own.

ScenarioRequests per monthTokens per request (in / out)Monthly cost
Hobby chat bot6001,500 / 300$0.225
Discord community9,0003,000 / 400$5.85
App with 1,000 users40,0002,500 / 500$25.00
Batch pipeline200,0001,200 / 250$61.00

The rest of the article shows the arithmetic and what drives each number.

Scenario 1: a hobby bot

You run a personal bot that answers about 20 messages a day, so 600 a month. Each message carries a short system prompt and some history, about 1,500 input tokens, and the reply averages 300 tokens.

  • Input: 600 x 1,500 = 0.9M tokens, which costs 0.9 x $0.15 = $0.135.
  • Output: 600 x 300 = 0.18M tokens, which costs 0.18 x $0.50 = $0.09.
  • Total: $0.225 per month.

This is the case where the minimum $20 top-up outlasts everything else. At this rate it would cover over seven years of use. Credit never expires, so a hobby project can sit on one top-up without a clock running.

Scenario 2: a Discord community bot

A bot in an active server handles 300 requests a day, or 9,000 a month. It reads recent channel context, so each call has about 3,000 input tokens, and it replies with around 400.

  • Input: 9,000 x 3,000 = 27M tokens, so 27 x $0.15 = $4.05.
  • Output: 9,000 x 400 = 3.6M tokens, so 3.6 x $0.50 = $1.80.
  • Total: $5.85 per month.

Watch the input side. Doubling the context you paste into each call doubles the larger part of the bill, while the reply length barely matters. Many bots save money by sending the last six messages instead of the last twenty. Also check the rate limits: a full key allows 8 requests at once and 300 per minute, which is far above this load, but a busy server can still spike when a thread takes off. Queue requests in the bot and you stay inside the limits.

Scenario 3: an app with 1,000 users

Say each of 1,000 monthly users sends 40 requests, which makes 40,000 requests. A request carries 2,500 input tokens (system prompt plus history) and returns 500 tokens.

  • Input: 40,000 x 2,500 = 100M tokens, so 100 x $0.15 = $15.00.
  • Output: 40,000 x 500 = 20M tokens, so 20 x $0.50 = $10.00.
  • Total: $25.00 per month, or 2.5 cents per user.

Here output is 40% of the bill even though it is only a sixth of the tokens, since it costs more than three times as much per token. Short answers pay off. If you asked for replies of 250 tokens instead of 500, the output line would fall to $5.00 and the total to $20.00. A $20 top-up would then cover this app for a month, and a $100 top-up, which comes with +10% bonus and so gives $110 of credit, would cover about five.

Scenario 4: a batch pipeline

Batch jobs scale by document count. Suppose you tag or rewrite 200,000 items in a month, each with 1,200 input tokens and a 250 token result.

  • Input: 200,000 x 1,200 = 240M tokens, so 240 x $0.15 = $36.00.
  • Output: 200,000 x 250 = 50M tokens, so 50 x $0.50 = $25.00.
  • Total: $61.00 per month.

For batch work the limit that matters is throughput, not price. At 8 concurrent requests, the job time is the number of requests divided by 8, times the average request duration. Run the arithmetic for your own latency before you plan the schedule, and keep your worker pool at 8 or fewer so you do not trigger the concurrency error. Failed requests are not charged, so retrying them adds no cost.

Estimate your own month

The formula is short: monthly cost = requests x (input tokens x $0.15 + output tokens x $0.50) / 1,000,000. This script does it and also reads real token counts if you run a sample first.

from openai import OpenAI

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

def monthly(requests, tokens_in, tokens_out):
    return requests * (tokens_in * 0.15 + tokens_out * 0.50) / 1_000_000

print(monthly(40_000, 2_500, 500))   # 25.0

# sample 20 real prompts, then use the measured averages
ins, outs = [], []
for p in open("prompts.txt").read().splitlines()[:20]:
    r = client.chat.completions.create(
        model="uncensored", messages=[{"role": "user", "content": p}])
    ins.append(r.usage.prompt_tokens)
    outs.append(r.usage.completion_tokens)
print(monthly(10_000, sum(ins) / len(ins), sum(outs) / len(outs)))

The sample run costs a fraction of a cent. New users can claim $0.50 of trial credit on the API key page to run it without paying anything, and the pricing page shows the rates and top-up bonuses.

What changes the bill most

Across all four scenarios, the same levers move the total:

  1. Context size. History and pasted documents drive input cost. Trim them.
  2. Output length. Output costs $0.50 per 1M, so ask for the length you need, and set max_tokens where it helps.
  3. Retries and loops. A bug that repeats calls can multiply a month's cost in an hour. Cap retries.
  4. Traffic growth. Cost scales linearly with requests. Double the users, double the bill.

Planning a budget

Pick the scenario nearest to your project and multiply by a safety factor of two for the first month, since real traffic is rarely as even as a spreadsheet. Then set a low-balance alert. Every response carries an X-Balance-USD header, so your app can email you when credit drops under a number you choose, and you top up before anything stops. If your traffic grows ten times, the bill grows ten times: the Discord bot at 10x would cost $58.50, and the app with 10,000 users $250.00. Nothing jumps in steps, so there are no pricing tiers to outgrow.

Funding it

You pay from prepaid credit with no subscription, so a quiet month costs close to nothing. Top-ups run from $20 to $500, by crypto (USDT on TRC20, USDC on Base), with +5% from $50, +10% from $100 and +15% from $200.

Limits of these estimates

Two things the scenarios leave out. First, they assume steady use, while real products have peaks, such as a launch day or a viral thread, and one peak day can use a week of budget. Second, they assume the prompts stay the same size. Product changes often grow the system prompt, and a 500 token addition to a call made 40,000 times a month adds 20M input tokens, or $3.00. Re-measure after every prompt change.

These are calculations, not promises about your usage. Real prompts vary, and your token counts depend on the language and the content you send. This is a text-only API, so costs for embeddings, images or audio are outside these figures. The service runs one open-weight model from an independent provider, tuned to answer without content refusals for lawful adult use.

For definitions of the terms used here, see what LLM inference is. Mistakes that quietly raise bills are listed under common AI chat API mistakes. The docs cover the request format.

Questions and answers

How much does a small bot cost per month?

A bot with 600 messages a month at 1,500 input and 300 output tokens costs $0.225.

Is there a monthly fee on top of token costs?

No. You pay per token from prepaid credit, and credit does not expire.

What is the smallest top-up?

$20, in any whole amount up to $500. Bonuses of 5%, 10% and 15% apply from $50, $100 and $200.

Do failed requests cost anything?

No. Failed or refused requests are not charged.

Your key is one form away

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

Get API key