Hosts compared

Where to run uncensored LLMs

The APIs and hosts that serve uncensored and abliterated models, with how they bill, what they say about logging, and how many of the ranked models each one serves. Logging is as each host declares it; we cannot audit it.

Host	Type	Billing	Logging	Crypto	Jurisdiction	Ranked models	Notes
RedPill	inference	per-token	Not stated	No	Not stated	2	Confidential inference in trusted execution environments (Phala), including an uncensored Qwen 3.8 27B.
Venice	inference	both	Says it keeps no prompts	Yes	United States	8	Private inference with its own uncensored model line and a crypto native API.
Featherless	subscription	both	Says it keeps no prompts	No	United States	–	Flat rate serverless access to thousands of Hugging Face fine tunes, 781 of them tagged uncensored.
ArliAI	subscription	subscription	Says it keeps no prompts	No	Not published	–	Unlimited token plans focused on creative writing and roleplay models.
NanoGPT	aggregator	per-token	Transient only	Yes	Not published	28	Pay per prompt aggregator. No account needed, crypto checkout, TEE hosted models.
OpenRouter	aggregator	per-token	Transient only	Yes	United States	–	General purpose router. Lists a handful of uncensored fine tunes alongside frontier models.
Abliteration.ai	inference	both	Says it keeps no prompts	No	United States	2	Hosted abliterated frontier models, led by Abliterated Large v2 on GLM-5.3.
Chutes	decentralized	per-token	Not stated	Yes	Decentralised	–	Decentralised GPU network on Bittensor with very low per token prices.
Open Uncensored	router	per-token, prepaid	Says it keeps no prompts	No	United States	31	One OpenAI compatible API in front of the hosts above, no account, provider price with no markup. Runs this site.

A dash means the host is not one Open Uncensored routes to, so its models are not in the ranking yet, not that it has none.