EN ▾
Get API key

LLM API Free Credits: How to Get and Use Them

Claim $0.50 of trial credit with one click, then see exactly what it buys and how to spread it over a week of useful tests.

Updated

Key points

  1. One click on the API key page adds $0.50; no card, one claim per person.
  2. The credit lasts 7 days and trial keys allow 2 concurrent requests and 60 per minute.
  3. $0.50 covers about 909 requests of 2,000 input and 500 output tokens.
  4. Output tokens cost more, so cap max_tokens in tests.
A glowing token feeding an hourglass beside seven tiles, for a seven-day trial

What you get and how to claim it

Most searches for llm api free credits end in a form, a waitlist or a card on file. Here the route is shorter. Sign up with Google, or with email and password, open the API key page and press one button to claim $0.50 of trial credit. There is no card step, and it is one claim per person.

The credit is valid for 7 days from the claim. Trial keys also run with lower limits: 2 concurrent requests and 60 requests per minute, against 8 and 300 on a full key. That is enough for testing and small scripts, and not enough for a production launch.

  1. Sign up with Google or email.
  2. Open the API key page and create a key.
  3. Press the trial button to add $0.50.
  4. Run your first request with the code below.

What $0.50 buys

Prices are $0.15 per 1M input tokens and $0.50 per 1M output tokens. The trial credit is $0.50, so it covers either 3.33M input tokens or exactly 1M output tokens, or any mix. Here is what that means for typical requests.

Request shapeCost per requestRequests for $0.50
1,000 in, 200 out$0.000252,000
2,000 in, 500 out$0.00055about 909
8,000 in, 1,000 out$0.0017about 294
50,000 in, 2,000 out$0.0085about 58

Check one row: 2,000 input tokens cost 2,000 x $0.15 / 1M = $0.0003. 500 output tokens cost 500 x $0.50 / 1M = $0.00025. The sum is $0.00055, and $0.50 / $0.00055 is about 909. The long-context row shows why big prompts burn credit fast: 50,000 input tokens alone cost $0.0075.

Failed or refused requests are not charged, so a request that errors does not use up any of it.

Your first request

Use any OpenAI SDK. Only the base URL, key and model name differ from the usual setup.

from openai import OpenAI

client = OpenAI(base_url="https://api.uncensoredendpoint.com/v1", api_key="YOUR_KEY")
r = client.chat.completions.create(
    model="uncensored",
    messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(r.choices[0].message.content)
print(r.usage)   # prompt_tokens, completion_tokens

Print usage from the start. Every response also carries an X-Balance-USD header, so you can watch the trial balance fall with each call. If a request fails, the header and the usage field still tell you what was and was not charged: failed or refused requests are free, so a failed call should leave the balance where it was. Check that on purpose once, by sending a request with a bad model name, so you know your error handling works before you rely on it.

curl -i https://api.uncensoredendpoint.com/v1/models -H "Authorization: Bearer YOUR_KEY"
curl https://api.uncensoredendpoint.com/v1/balance -H "Authorization: Bearer YOUR_KEY"

A seven-day test plan

Seven days and about 900 mid-sized requests go a long way if you plan them. A split that works for most projects:

  • Day 1, about $0.02. Smoke tests: one chat, one streamed chat, one request with tools, one with JSON mode. Confirm each works with your client library.
  • Days 2 to 3, about $0.10. Quality test. Run 30 to 50 real prompts from your product and read every answer.
  • Days 4 to 5, about $0.10. Behaviour test. Check how the model handles your system prompt, long conversations and the topics your users will raise.
  • Days 6 to 7, the rest. Load test at the trial limits: 2 at once, 60 per minute. See what a retry wrapper has to do.

Keep a small log of prompt, answer, token counts and cost. At the end you will have an average cost per request, which lets you project a monthly bill before you spend anything.

Make the credit last

Output tokens cost over three times as much as input tokens, so limit them first. Set max_tokens on tests that only need a short answer, and ask for brief replies in the system prompt. Next, keep prompts lean: send only the conversation turns you need, not the whole history.

Use a fixed seed and low temperature when comparing prompt versions, so differences come from your change and not from sampling noise. Test with short inputs first, and move to long ones only after the short ones pass. A single 90,000 token document costs $0.0135 of input, about the same as 54 small chats.

Avoid loops that call the API with no cap. A bug that retries every second can drain the credit in minutes, so put a counter and a hard stop in any script you leave running.

Track spend while you test

A tiny wrapper that records tokens and cost per call gives you real numbers instead of guesses. The constants below are the published prices per token.

import csv, time
from openai import OpenAI

client = OpenAI(base_url="https://api.uncensoredendpoint.com/v1", api_key="YOUR_KEY")
IN_PRICE, OUT_PRICE = 0.15 / 1_000_000, 0.50 / 1_000_000

def run(prompt, label):
    t = time.time()
    r = client.chat.completions.create(
        model="uncensored",
        messages=[{"role": "user", "content": prompt}],
        max_tokens=300,
    )
    u = r.usage
    cost = u.prompt_tokens * IN_PRICE + u.completion_tokens * OUT_PRICE
    with open("trial_log.csv", "a", newline="") as f:
        csv.writer(f).writerow([label, u.prompt_tokens, u.completion_tokens,
                                round(cost, 6), round(time.time() - t, 2)])
    return r.choices[0].message.content

After a day of use, sum the cost column and divide by the number of rows. That average is the figure to carry into your budget. Compare it with the table above: if your real prompts are longer than you assumed, you will see it in the first hour, not on your first invoice.

How to judge any free-credit offer

The same questions apply to every provider, so ask them before you build around a free tier. Does it need a card? How long does it last? What are the rate limits on a trial key, and do they match your test plan? Is the credit tied to one model, or can you try the one you intend to ship? Check each provider's current terms, because free offers change often.

One more question people forget: what happens to the integration afterwards? An OpenAI-compatible endpoint means the code you write during the trial moves to a paid plan, or to a different host, by changing a base URL. That makes a short trial low-risk even if you decide to go elsewhere.

What the trial does not cover

Be clear about the edges. The trial is for testing, and it ends after 7 days whether or not you used all of it. It does not raise the limits, and you cannot claim it twice. It also does not change what the service is: one open-weight model, run by an independent provider, tuned to answer without content refusals for lawful adult use. It is text only, with no images, audio, embeddings or fine-tuning, and sexual content involving minors is never allowed.

Prompts are not used for training. Still, do not paste secrets or private customer data into any test, free or paid. Test with synthetic data first.

After the trial: paying for more

When the credit runs out, you move to prepaid credit with no subscription. Any whole amount from $20 to $500 works, paid by crypto (USDT on TRC20, USDC on Base). Larger top-ups add a bonus: +5% from $50, +10% from $100 and +15% from $200, so $100 gives $110 in credit and $200 gives $230. Credit never expires.

At the 2,000 in, 500 out shape, a $20 top-up is about 36,000 requests. See the pricing page for details, the endpoint pricing notes for worked costs, and the docs for every parameter.

Questions and answers

Do I need a card for the free credit?

No. You sign up with Google or email and claim the $0.50 with one click on the API key page.

How long does the trial credit last?

It is valid for 7 days from the claim. Unused credit after that is gone.

Can I claim it more than once?

No. The trial is one per person.

Is paid credit different from trial credit?

Paid credit never expires and the key gets full limits of 8 concurrent requests and 300 per minute.

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