Uncensored and Abliterated
Two approaches to following your requests, with different refusal rates.
Abliterated: 0% refusal rate
Weight changes remove the behavior that causes a request to be rejected. In our catalogue, this category includes GLM 5.3, GLM 5.3 Flash and DeepSeek V4.1 Flash.
Uncensored: up to 10%
Kimi 3.1P, Qwen 3.8 Max and DeepSeek V4 Pro offer more freedom for writing, coding and open-ended generation. Some requests may still be refused.
These versions may use prompting, fine-tuning, weight changes or a combination of techniques.
Reading the percentages
The refusal rate is the share of requests a system declines to answer. These figures are stated by Uncensured AI and refer to the versions offered through the service.
Models available on Uncensured AI
The list below follows the current catalog. Open a model page for its API ID and current token prices.
- Kimi 3.1P Uncensured AI
kimi-k3.1-pro-uncensored - DeepSeek V4.1 Flash Uncensured AI
deepseek-v4.1-flash-abliterated - GLM 5.3 Flash Uncensured AI
glm-5.3-flash-abliterated - GLM 5.3 Uncensured AI
glm-5.3-abliterated - DeepSeek V4 Pro Uncensured AI
deepseek-v4-pro-uncensored - Qwen 3.8 Max Uncensured AI
qwen3.8-max-uncensored