Cleus
AI that answers the question instead of refusing it
Start new thread

Cleus - AI that answers the question instead of refusing it

by•
Cleus answers the questions other AI refuses. Chat with web search, code, images and video, free to start. Cleus Code, a coding agent that never stops at "I can't help with that". New: DeepSeek, GLM, Qwen and Kimi with the refusals removed, on one API key.

Add a comment

Replies

Best
Maker
📌

Hi Product Hunt, I'm Bucky, the founder of Cleus.

Cleus is uncensored AI: it answers the question you asked, without the lecture. It now comes in three forms, and the third is new today.

1. Cleus, the chat app

An uncensored agent in your browser or on your phone. It searches the live web, runs real code, reads your files, and makes images, video, characters and long stories from the same account. It is free to start, with no card.

2. Cleus Code, the coding agent

A desktop and terminal coding agent for Mac and Windows. It writes and runs code, drives your browser and desktop apps, connects to GitHub and any MCP server, and can make images, video and voice for your project. It runs on Cleus V3, so it does not stop at "I can't help with that", and other models are one setting away.

3. Cleus API, new today: abliterated open models

Five open-weight models with the refusals removed, each under its own name, behind one OpenAI-compatible key:

  • deepseek-v4.1-flash-abliterated

  • glm-5.3-abliterated

  • glm-5.3-flash-abliterated

  • qwen-3.8-27b-abliterated

  • kimi-k3.1-pro-abliterated

"Abliterated" means the model's habit of refusing, which shows up as a direction inside its activations, is removed from the weights. Nothing is retrained and no jailbreak prompt is involved.

These five are a direct line: system, user, assistant and tool messages reach the model in the roles you gave them, and we add no persona. The one exception is Kimi, which gets a single short note saying which model it is and that Cleus provides it. Change the base URL to and existing OpenAI code works, with streaming, tool calls and reasoning_effort. For these models we record token counts, cost and time for billing, and we do not save prompts or replies.

Cleus V3, our own uncensored model and the one behind the chat app and Cleus Code, is on the same key. It covers deep reasoning, fast replies and coding, and it answers as Cleus.

The honest parts

  • Chat is free to start. Cleus Code needs Cleus Premium plus a one-time $99 add-on, and all $99 goes to your balance. The API needs Premium plus a prepaid balance. There is no free API tier.

  • Thinking models can be slow to start. I have seen a first token take over 30 seconds on a bad run.

  • Kimi is the newest of the five and the least steady. Tell me if a reply comes back empty.

I'm here all day. Tell me what you tried, which model you want added next, and what broke.