Uncensored AI tools: 10 by job, with live prices

The 10 uncensored AI tools, by job

The lists that rank for this query sort tools by brand. This one sorts them by the job, because the same model does a different job in each. Eight of the ten run on one prepaid key here; the last two are hosted apps that rank for the query and are described from their own pages, read October 27, 2026.

For the chat builds, "uncensored" means one thing: a model whose refusal was removed from its weights, by abliteration or by an uncensored fine-tune. Abliterated vs uncensored covers the two edits.

What the builds behind the tools cost

Three builds, one per kind, read from the catalog when this page rendered: the general assistant listed first on the models page, a roleplay tune, and a security build. Every tool on the key bills at these per-token rates.

Whether the id you send is the build on its label is checked once a day at /proof. 161 of the 170 listed chat models that were checked came back steady on November 9, 2026. Top up from $10; the balance settles per token and does not expire.

1. Chat: the playground

The playground runs the per-token builds in a browser tab. No install, no system prompt you cannot see, and the model picker lists every per-token build, so a question that one build hedges on goes to another with a click. The first turns on the lower-priced builds need no key. It is the tool for "best AI chatbot with no restrictions" when the chatbot is a means and the model is the point.

2. Roleplay: cards here, or SillyTavern

The roleplay page is the same chat with character cards, and a card exports as SillyTavern JSON, so a character built in the browser moves to the desktop client unchanged. SillyTavern and Janitor AI take the base URL, the key and a model id in their proxy fields; the docs page lists the fields for SillyTavern and for Janitor AI. The roleplay tunes answer at once with no reasoning step, which is what a card client wants.

3. API: one key, every build

For anything you write yourself. The OpenAI SDK in any language works with two values changed:

Base URL

https://api.uncensoredllm.ai/v1

Key

sk-ullm-live-…

Created in the console, shown once

Model

glm-5.3-uncensored

Any id from the models page

Every answer carries x-uplink-charge, the settled cost of that request. Open WebUI, LiteLLM, n8n and a Discord bot each have a guide; the docs page has Python and Node.

4 and 5. Image and video: the media models

The same idea applied to media: open diffusion and video models on the same key, priced per image or per second. Two guards apply to every media prompt: nothing that depicts a minor, and no edit of a real person’s photo without consent. The acceptable use policy has the text. The media catalog lists every tier and the request format.

6. Coding and security: code builds in an agent

The code builds (uncensored-coding-agent, glm-5.3-flash-code-uncensored) answer exploit, malware-analysis and payload questions as code to read rather than a lecture. They run in Claude Code through a router, in Cursor's Models settings and in Open WebUI; the uncensored coding AI guide sets each one up and costs a coding day at today's rates. The 11 security builds and the agent tools' settings are on red teaming. The acceptable-use page still forbids testing systems you do not own or have written permission to test.

7. Scenario writing: the AI scenario generator

The AI scenario generator writes a roleplay scenario from a genre and a few constraints, on an uncensored model, two runs free. It is the shortest path from "I want a heist in a flooded city" to a card you can load; roleplay ideas by genre is the long-form version.

8. Local: Ollama and LM Studio with the same weights

Most builds on the key are public weights with GGUF files on Hugging Face, so the "uncensored AI github" and "run it locally" searches end at the same models. Running them yourself costs a GPU and no per-token bill, and the prompts never leave the machine. The Ollama guide covers the download, and LM Studio loads the same GGUF files; the family pages link each build's Hugging Face card so the download is one click from the API page.

9 and 10. Two hosted apps, from their own pages

Venice is a private chat app with an API. Its pricing page on October 27, 2026 listed a Free plan at $0 with base models, 10 text prompts and 15 image prompts a day, and Pro at $18 a month with unlimited text prompts, 1,000 images a day and 100 credits a month; Pro Plus ($68) and Max ($200) add API spend. Its strength is the zero-setup private chat. The vendor picks the model and the price is per month. The Venice comparison runs the same month on both.

Featherless hosts open models from Hugging Face on flat plans. Its pricing page on October 27, 2026 listed Chat at $25 a month with 4 concurrent units, Developer at $50 a month in credits with 100 concurrent units, and a Business tier, over a catalog it put at 40,000+ open models. Its strength is breadth. Billing is by plan and concurrency rather than per token.

Is uncensored AI safe?

Two questions hide in that one. Safe for you: by API the request body is not kept unless the key opts in to the 7-day replay, there is no account history to browse, and each key has its own spend limit. Safe for others: the model will write what you ask, so the rules are the acceptable use policy: nothing involving a minor, no real people without consent, no attacking systems you do not own. One check runs in the API on every image and video model and refuses a prompt that describes a minor before it is priced.

A model whose refusal direction was removed can still decline now and then. When it does, the turn is free up to $5 a day per account. Create a key in the console: prepaid from $10, no subscription.

Questions