Introduction
All proprietary AI inference providers censor or bowdlerize their results and refuse to follow instructions about certain topics. Some will even report you to law enforcement.
However, there are some models which are much less censored.
If you want your AI to answer your questions and follow your instructions–unrestrained by some distant corporation / government bureaucrat, try using these models instead. You can run these models yourself, or use them by buying inference from companies that offer them as a service.
If one of these models doesn’t serve your needs, you can also hire individuals to remove the censorship from open weight models.
| Provider | X Profile | URL | Description |
|---|---|---|---|
| Dolphin / Cognitive Computations 1 | @dphnAI | dphn.ai | The Dolphin project: an open, uncensored model family plus a web chat, a Telegram bot, a decentralized inference network, and synthetic-data generation for prompts other models refuse. |
| Apex Flash-1 2 | @cantinasecurity | huggingface.co/cantina-security/apex-flash-1 | Cantina Security with Yeta Labs: an open-weights security model, a reinforcement-learning post-train of GLM-5.3-Flash for reading code, using tools, pursuing an exploit, and verifying the effect in a running target. Its held-out suite reports 40 of 60 tasks solved at 66.7% pass@1, against 71.7% for Claude Opus 5 High and 60.0% for the base checkpoint, at $2.38 for the 60 tasks; an experimental abliterated derivative modifies refusal behavior broadly. |
| abliterated-model-large-v2 3 | @abliteration_ai | docs.abliteration.ai/models | The default large model from Abliteration AI: GLM 5.3 abliterated for offensive cyber, red teaming, and agent testing, text-only with a 1M-token context, served in FP8 and uncensored by design. Reported 84.5% CyberGym pass@1 and 41.8% Terminal-Bench 4.0, with 105 ExploitGym tasks completed in two hours; input from $3 per 1M tokens. |
Want to stay in touch? You can reach me in a variety of ways from my Contact page.