Unrestricted AI chat

Short answer: a refusal is trained behaviour, and it can be trained back out. Uncensored models on the Pro plan ($45 a month with $35 of usage) answer where mainstream assistants stop, and the account they need is a random number rather than an email.

The refusal is not the model being stupid. Assistants are tuned to be safe in the general case, so a question that is slightly unusual gets the same treatment as a hard one. Ask a mainstream model to quote a paragraph of a song, summarise a paywalled article, or explain a bug in code that belongs to a closed product, and you get the disclaimer voice rather than the answer.

What an unrestricted model actually is

Two techniques, both of them maths rather than policy:

Neither makes a model omniscient. Both make it much less likely to open with “I can’t help with that”, so you spend your turn on the question instead of on persuading it.

Models that refuse least, right now

Read live from the catalogue. Each page has the price, the context length and an API snippet.

These are open models, so there is no provider deciding on your behalf. They are on Pro, whether you use the web chat or the API.

Where the limit actually is

Uncensored is a spectrum, and anyone claiming the end of it is overselling. Large frontier models still reason better on hard problems; an uncensored 70B will happily attempt something a frontier model would hedge on, and then get the details slightly wrong. The useful pattern is to use both: unrestricted for the first pass and the awkward material, a frontier model on the Casual plan ($20 a month) when correctness matters more than permission.

The Terms of Service still apply and you are still liable for what you do with the output. Uncensored also does not mean “evades law enforcement”, whatever the marketing says.

No email, no login

The account uncensored models need is a random 16-digit number. No email, username, password or phone. Pay with Monero, Bitcoin, Lightning or a card, all of which become blind vouchers, so the payment is not attached to the account it funds.

For a first look with nothing at all, guest chat runs Llama 3.3 70b Instruct with 30 messages an hour, no account and no cookies. That is a standard model, not an uncensored one.

Common questions

Which models are uncensored here?

The live list is above. It changes as models come and go, which is why it is read from the catalogue instead of written into this page.

Do uncensored models need a paid plan?

Yes, Pro at $45 a month with $35 of usage included. They are the most expensive thing to serve, so they are the one category that stays behind a plan.

Can I use them with no login?

Not with no login at all. With no identity: yes. The account is a number.

Is an abliterated model worse at normal tasks?

Somewhat, on reasoning-heavy work. Abliteration costs a little accuracy. Use a frontier model when the answer has to be right.

Can I use these through the API?

Yes, with the same Pro plan and the same model ids, on the OpenAI-compatible endpoint at /v1.

What about images?

Text only. There is no image generation here, so “unrestricted image generator” comparisons will not flatter us.

Last updated 2026-11-02. Plan prices and model lists on this page are read live from the site.

Try it now: guest chat needs no account, no email and no cookies. When you want more models, an account is just a random number.

Try an unrestricted model See pricing
Try an unrestricted model