Blog

How uncensored models work, how we compare, and what is new.

Guide · October 28, 2026What is an abliterated model?Abliteration removes the refusal behavior from an open-weight LLM without retraining it. How it works, what it costs in quality, and how it compares to uncensored fine-tunes and jailbreaks.
Comparison · October 28, 2026Venice AI API alternative for uncensored modelsSilk Compute and the Venice AI API compared for uncensored LLM work: models, context, prices, API formats and payments. When each one is the better fit.
Comparison · October 28, 2026Uncensored models on OpenRouter vs a dedicated uncensored APIWhy routing through OpenRouter does not remove refusals, what to look for in uncensored models, and how to use Silk Compute next to OpenRouter.
Launch · October 26, 2026Kimi K3.1 Uncensored is liveAbliterated Kimi K3.1 is now on Silk Compute: our flagship uncensored model for agentic coding and research, with 1M tokens of context and up to 944K output tokens.
Launch · October 26, 2026DeepSeek V4.1 Flash 1017 Uncensored is liveAbliterated DeepSeek V4.1 Flash 1017 is now on Silk Compute: fast reasoning with low, high or max effort, 1M tokens of context, and no refusals.
Launch · October 26, 2026GLM-5.3 Flash Uncensored is liveAbliterated GLM-5.3 Flash is now on Silk Compute: our lowest-cost uncensored model, at $0.30 input and $1.00 output per 1M tokens, with 1M tokens of context.