AI API Free Tier: What Developers Should Expect
An ai api free tier is meant to get you to a first working request, not to run a product. Here is what to expect from one, what to check before you build on it, and how a small trial plus pay-per-token credit works instead.


Key points
- Free tiers are for evaluation: small quotas, tight rate limits, sometimes a card on file.
- Check the quota, the model list, the rate limits and what happens at the cap.
- Here: $0.50 trial, one click, no card, 7 days, then prepaid credit from $20.
- At $0.15 and $0.50 per 1M tokens, $20 pays for tens of thousands of chat messages.
What a free tier is for
Providers offer free access for one reason: to let you try the API before you commit. That shapes the terms. A free tier is usually small, limited in speed, and sometimes limited in which models you can call. Some require a card up front, some expire, some are for one project only. These details change often, so read each provider's current terms instead of trusting a blog post, including this one.
For a developer the useful question is not "is it free?" but "does it let me finish the evaluation?" A prototype needs enough requests to test your prompts, enough speed to test streaming, and enough access to try the model you plan to ship with. If the free tier fails any of those, you are testing the limits instead of the product.
Five things to check before you build on one
- Quota type. Is it a dollar credit, a token count, or a request count? Credits and tokens are easy to budget. Request counts hide how long your prompts are.
- Rate limits. Low requests per minute and low concurrency are normal on free plans. Check whether streaming holds a slot while it runs.
- Model access. A free tier may offer a smaller model than the one in the paid plan. Then your test results do not carry over.
- Expiry. Credit that disappears in a week is fine if you plan for it and a nuisance if you do not.
- The upgrade path. What does the first paid month look like? Subscriptions, minimum commitments and tier jumps change your cost model.
Go through the list for every provider you compare. It takes ten minutes and saves a rewrite later.
How the trial here works
This service does not run a permanent free plan. New users can claim $0.50 of trial credit with one click on the API key page after signing up with Google or with email and password. No card is needed, the credit is valid for 7 days, and it is one per person. The free trial FAQ answers the usual follow-up questions.
The trial key has the same API as a paid one, with lower limits: 2 concurrent requests and 60 requests per minute, against 8 and 300 for a full key. You call the same model, uncensored, an open-weight model run by an independent provider and tuned to answer without content refusals for lawful adult use. It has a 100,000-token context window, streaming, tool calling and JSON mode. It is text only.
That means what you learn in the trial applies directly to production, apart from the limits. Failed or refused requests are not charged, so errors during testing do not eat the credit.
What $0.50 buys, and what $20 buys
Pricing is $0.15 per 1M input tokens and $0.50 per 1M output tokens, pay as you go. A typical chat message in a prototype might be 1,500 input tokens (system prompt plus history) and 300 output tokens.
| Part | Math | Cost |
|---|---|---|
| Input | 1,500 x $0.15 / 1,000,000 | $0.000225 |
| Output | 300 x $0.50 / 1,000,000 | $0.00015 |
| One message | sum | $0.000375 |
| 1,000 messages | 1,000 x $0.000375 | $0.375 |
So the $0.50 trial covers about 1,333 such messages. That is enough to test a feature thoroughly, and not enough to run a beta with real users. The smallest paid top-up is $20, which covers about 53,333 messages at the same size. Top-ups can be any whole amount from $20 to $500, with +5% bonus credit from $50, +10% from $100 and +15% from $200. A $100 top-up, for example, becomes $110 of credit. Credit never expires and there is no subscription. Full details are on the pricing page.
Free tier or prepaid credit: which fits which job
Neither is better in general. They suit different stages.
| Situation | Better fit |
|---|---|
| Checking that an SDK connects and the response format works | Any free tier or the $0.50 trial |
| Comparing how models answer your prompts | Whatever gives access to the model you would ship |
| A demo that must stay up for weeks | Prepaid credit, since a free quota can run dry mid-demo |
| Load testing or a closed beta | Paid, with full limits |
| A hobby project with a few messages a day | A small top-up goes a long way at these prices |
If you are already using the OpenAI SDK, switching costs almost nothing, because you only change the base URL and key. The page on using an OpenAI-compatible API shows the change in a few lines.
A prototype plan that fits the trial
Spend the trial like a budget. This split works for most projects:
- Day 1, about $0.05: connect, list models, send ten varied prompts, check streaming.
- Day 2 to 3, about $0.20: run your real prompts from a spreadsheet through a loop and read every answer.
- Day 4 to 5, about $0.10: test tools, JSON output and error handling.
- Keep $0.15 spare for the checks you forgot.
A simple loop is enough:
from openai import OpenAI
client = OpenAI(base_url="https://api.uncensoredaiapis.com/v1", api_key="sk-your-key")
prompts = ["Summarize: ...", "Rewrite this politely: ...", "Extract the dates from: ..."]
for p in prompts:
r = client.chat.completions.create(
model="uncensored",
messages=[{"role": "user", "content": p}],
)
print(r.usage.total_tokens, r.choices[0].message.content[:80])
The code examples page has more ready-to-run snippets, including streaming and tools.
Signs a free tier is the wrong tool
Move on to paid credit when you notice any of these: you are waiting on rate limits more than on answers, your tests keep failing at the quota and not at the code, you need a larger context window than the free model offers, or you are about to show the product to real users. At that point the question changes from "can I try it?" to "what will it cost per month?", and the answer is easy to compute from the per-token prices above. Take your average message size, multiply by expected messages, and add a margin. For example, 2,000 users sending 10 messages a month at $0.000375 each is 20,000 messages, or $7.50 a month. Double the message size and you double the bill, which is why trimming history and capping long replies matters more than any discount. Run the same arithmetic for each provider you compare, using their current published prices, and the decision usually becomes obvious. If you are leaving another service, the guide to migrating from OpenAI or OpenRouter covers the switch.
More guides
Questions and answers
Is there a permanent free plan?
No. New users claim a one-time $0.50 trial valid for 7 days. After that you add prepaid credit, from $20.
Do I need a card for the trial?
No card is needed. Sign up with Google or email and password and press the trial button on the API key page.
Can I use the trial credit for any request?
Yes, it works on the same API as paid credit, with the trial limits of 2 concurrent requests and 60 requests per minute.
Are my prompts used for training?
No. Prompts are not used for training.
Your key is one form away
Create an account, copy the key, change the base URL. That is the whole setup.