NoFilterGPT is a “no censorship” chat app with image and video tools and a REST API. It runs one custom model, nofilter-large, and sells access by subscription rather than by token. Its marketing promises “without limits, logs, blocks, no exceptions”. Its own terms and privacy policy are more qualified.
This review is based on NoFilterGPT’s published pricing, model information, and policies, checked on October 11, 2026. Benchmark results from our own runs will be added when they’re complete. Live numbers are on the NoFilterGPT provider page.
The short version
| Best for | A cheap flat-rate chat subscription with some API access included |
| Uncensored models | 1 (undisclosed method) |
| Per-token price | Not published; API access comes with the subscription |
| Plans | Free (7 messages a day), Professional $11.99/mo |
| Privacy score | 32 / 100, 23rd of 35 |
| Watch out for | No legal entity or address, and an unnamed upstream vendor in the terms |
Models
We list one model: nofilter-large. NoFilterGPT’s FAQ calls it “a custom language model based on the GPT architecture but without the classic restrictions”. That doesn’t identify the base model, the size, or the uncensoring method. Its terms mention a third-party AI engine, and OpenRouter’s public stats show NoFilterGPT sending daily traffic to ReMM SLERP 13B, a Llama 2 13B fine-tune that Mancer serves. So we list NoFilterGPT as an aggregator, not a model host. We mark the model “Unverified” until we test it.
The terms also still ban several categories of use, despite the “no exceptions” marketing.
Search it on the Models page.
Pricing
There’s no per-token price. You choose a plan:
- Free: 7 messages a day, with limited image generation.
- Professional, $11.99/mo: “near unlimited” chat, image tools, and API access capped at 750 API messages a day.
If you used the whole API allowance every day, that’s about 22,500 messages a month, or roughly $0.53 per 1,000 messages. That’s competitive with per-token pricing for short messages. But you can’t buy more API capacity than the cap, and the docs don’t state OpenAI compatibility, so you may need to write against its own REST format.
Privacy
NoFilterGPT scores 32 / 100 on our privacy rubric, which only gives points for documented claims:
| Rubric item | NoFilterGPT | Points |
|---|---|---|
| Prompt retention | Says conversations are purged after being received; policy logs IPs and usage data | 10 / 20 |
| No training on prompts | Not stated | 0 / 15 |
| Technical guarantee (TEE / E2EE) | None; policy only | 0 / 20 |
| Signup | 5 / 8 | |
| Crypto payments | BTC, ETH, USDC | 7 / 7 |
| Jurisdiction | US governing law | 10 / 10 |
| Upstream hosts disclosed | No; terms mention an unnamed vendor | 0 / 10 |
| Independent audit | No | 0 / 10 |
The homepage says “We do not keep conversations” and that interactions “are purged after being received”. The privacy policy says it logs IP addresses and usage data, without a stated retention period. We give partial credit for retention because the content claim is clear but the metadata logging is open-ended.
The bigger gap is the upstream vendor. If prompts go to a third party, that party’s logging policy matters too, and NoFilterGPT doesn’t name it. The site also lists no legal entity or postal address, only an email.
Access and payment
- Web chat, image and video tools, and a REST API on the paid plan.
- Email signup.
- Card, Google Pay and Apple Pay, or crypto (BTC, ETH, USDC).
How it compares
- Named company, similar price: Venice.ai (privacy 55) has 10 free prompts a day, Pro at $18/mo, and 16 named uncensored models.
- Flat-rate with more models: Infermatic (40) is $9/mo with 9 uncensored models and a named Delaware jurisdiction.
- Pay per token, no email: AnonRouter (80) lists 9 named uncensored models from $0.07 / $0.40 per 1M tokens and names the host behind each request.
Verdict
NoFilterGPT is inexpensive and accepts crypto. The flat $11.99 plan with a daily API allowance could suit someone who wants chat plus a light API without tracking token spend.
But you’re trusting an operator with no published name or address, running a model it doesn’t identify, possibly through a vendor it doesn’t name. The “no logs” headline is also softer in the actual policy. If you care where your prompts go, a provider with a named company and model list is a safer choice.
See NoFilterGPT’s live details on its provider page, or compare every provider in the index.