What free means here
Unrestricted ai free sounds like an unlimited chatbot with no price tag. For an API, that model does not hold up: every token costs GPU time, and someone has to cover it. So the free part here is a trial, sized for testing rather than production.
The model itself is one open-weight text model, run by an independent provider and tuned to answer without content refusals for lawful adult use. "Unrestricted" describes refusal behaviour, not a lack of rules: sexual content involving minors is always blocked, and the service is text only.
If you want a quick explanation of the trial mechanics, the free trial FAQ answers the common questions. This article focuses on what the credit buys and when it stops being enough.
How to claim the trial
- Sign up with Google, or with email and password.
- Open the API key page.
- Press the trial button. $0.50 of credit lands on the account.
- Copy your key and call the API.
It is one trial per person, there is no card to enter, and the credit is valid for 7 days. Those 7 days start when you claim, so claim it when you are ready to test, not the moment you sign up.
Fair limits on a trial key
Trial keys are deliberately smaller than full keys:
| Trial key | Full key | |
|---|---|---|
| Concurrent requests | 2 | 8 |
| Requests per minute | 60 | 300 |
| Context window | 100,000 tokens | 100,000 tokens |
| Features | Streaming, tools, JSON mode | Streaming, tools, JSON mode |
Everything works on the trial key. The only differences are the two rate limits. If you go over, the API returns HTTP 429 with the code rate_limited (per minute) or concurrency (too many at once). Wait a moment and retry; nothing is charged for a rejected request.
What $0.50 actually covers
Pricing is $0.15 per 1M input tokens and $0.50 per 1M output tokens. A few examples, worked out exactly:
- Short chat message: 300 input and 400 output tokens. Input: 300 x $0.15 / 1M = $0.000045. Output: 400 x $0.50 / 1M = $0.0002. Total $0.000245, so $0.50 covers about 2,040 messages.
- Long story scene: 2,000 input and 3,000 output tokens. Input $0.0003, output $0.0015, total $0.0018. That gives about 277 scenes.
- Big context test: 90,000 input and 2,000 output tokens. Input $0.0135, output $0.001, total $0.0145. That gives about 34 near-full-window calls.
In other words, the trial lets you run hundreds of realistic prompts. It does not let you run a public app for a week, and it is not meant to.
A good first test
Use a handful of calls to answer the questions that matter. Does the model follow your persona or system prompt? Does it refuse where you need it not to? Does tool calling return valid JSON? A minimal script covers the first one:
from openai import OpenAI
client = OpenAI(base_url="https://api.unrestrictedaiapi.com/v1", api_key="YOUR_KEY")
resp = client.chat.completions.create(
model="uncensored",
messages=[
{"role": "system", "content": "You are a blunt noir narrator. Stay in character."},
{"role": "user", "content": "Describe a rainy street in four sentences."},
],
temperature=0.9,
)
print(resp.choices[0].message.content)
print(resp.usage)
The usage block shows exact token counts, which you can plug into the maths above. For more languages, see the quickstart in five languages, and for prompt style with permissive models, read prompting unrestricted models.
When to move beyond the trial
Top up when the trial runs out or expires, whichever comes first. Any whole amount from $20 to $500 works, and credit never expires. Bonuses apply on larger top-ups: +5% from $50, +10% from $100 and +15% from $200. You can pay with crypto (USDT on TRC20, USDC on Base).
A $20 top-up at the short-chat cost above covers roughly 81,600 messages ($20 / $0.000245). Even if your messages are ten times longer, that is still thousands of conversations. The pricing page has the full breakdown, including exact bonus tiers.
Before you commit, remember what the service is. It is one model, ID uncensored, and not any vendor's official API. It handles text only, with no images, audio, video, embeddings or fine-tuning, and it is meant for lawful adult use. Test your own use case during the trial rather than assuming it will fit. The unrestricted API overview covers the product in more detail.
You also get a full key's limits: 8 concurrent requests and 300 per minute. Prompts are not used for training.
Why most free unrestricted chatbots disappoint
Free web chatbots that promise no filters tend to share a few traits. Understanding them helps you decide whether a trial-then-pay API suits you better.
- Queues and caps. Free capacity is shared, so you wait, or hit a daily message limit that resets whenever the operator decides.
- No control. You cannot set a system prompt, temperature or stop sequence, so characters drift and replies cannot be tuned.
- No memory you own. History lives on someone else's server, in a format you cannot export or feed into your own tools.
- Unclear data use. Free services often use conversations to improve their product. Here, prompts are not used for training.
An API flips each of these. You decide the system prompt, you store the history, you choose when to spend. The price of that control is a few cents, and the trial lets you check it first without paying anything.
Making the trial last
The credit is small, so a few habits stretch it a long way:
- Set
max_tokensto 300 to 500 while you are testing prompts. Output tokens cost more than three times as much as input. - Test with short histories first, then run one long-context test at the end.
- Read the
X-Balance-USDresponse header after each call to see the real cost, instead of estimating. - Put your test prompts in a file and re-run the same set after every change, so you compare like with like.
A reasonable plan is 40 to 60 calls over the first day, a review of cost and quality, and a decision before the 7 days end. If you plan a Discord or Telegram bot, remember that every user message in a shared channel spends credit, so keep a per-user message limit in your own code from the first day. Without one, a handful of enthusiastic testers can drain the trial in an afternoon, and the 60 per minute cap will start rejecting requests long before the money runs out.
