Frequently asked questions
Short questions, direct answers. If yours is not here, ask on Telegram.
01What is UncensoredAI?+
An OpenAI-compatible API gateway. One key and one /v1 endpoint reach GPT, Claude, Gemini, Grok, DeepSeek, GLM, MiMo and Kimi — including uncensored and abliterated variants — billed per token from a prepaid crypto balance.
02Do I need an email or KYC to get an API key?+
No. There is no email, phone number, password or identity check. One button mints a random 16-digit key, which is the account and the API credential.
03Is the API really OpenAI-compatible?+
Yes. Point any OpenAI SDK at https://uncensoredai.dev/v1, set the key as the Bearer token, and keep your existing request, streaming and error-handling code unchanged.
04What does uncensored mean?+
The provider's safety refusals are relaxed on a specific model id. The model runs as a normal catalog entry on the same endpoint, so you can compare it against the standard version side by side.
05What does abliterated mean?+
The refusal direction is removed from the model's weights. The change is baked into the model rather than applied by a system prompt, so it persists across every request.
06Which uncensored and abliterated models are available?+
GLM-5.3 Uncensored, GLM-5.3 Flash Uncensored, GLM-5.3 Abliterated, GLM-5.3 Flash Abliterated and DeepSeek V4 Pro Uncensored — listed alongside every frontier model in the same catalog.
07How is pricing calculated?+
Per 1,000,000 tokens against your prepaid balance, using the per-model input, cached and output rates on the pricing page. There is no subscription, minimum or seat fee.
08Which model is the cheapest?+
DeepSeek V4.1 Flash, at $0.04 per 1M input tokens. MiMo V2.6 Flash and GLM-5.3 Flash are the other low-rate models for high-throughput work.
09Do cached tokens cost less?+
Yes. Repeated prompt prefixes served from cache bill at the discounted cached rate shown per model — typically a fraction of the input rate.
10Does the API support streaming?+
Yes. Set stream: true and the response is text/event-stream, one delta per data: line, ending with data: [DONE]. The final frame may carry a usage object.
11Does the API support function calling and tools?+
Yes. The standard tools array is accepted and forwarded upstream. Tool support depends on the model; use a reasoning tier such as GPT-6 Sol or Claude Opus 5.5 for agent loops.
12How do crypto deposits work?+
Choose an amount in Billing and complete the hosted invoice in a supported cryptocurrency through Heleket. A signed webhook credits the exact USD amount once. Replayed webhooks cannot double-credit the balance.
13Are there rate limits or spending caps?+
Requests are rate-limited per key. Spend is naturally capped by your prepaid balance — when it reaches zero, requests fail with 402 insufficient_quota until you top up.
14How do I rotate or revoke my key?+
Rotate from the dashboard's Access key page. The old value is invalidated immediately and other sessions are signed out. We store only a hash, so a lost key cannot be recovered — only replaced.
15Is using uncensored models legal?+
That depends on your jurisdiction and your use. The models are tools; responsibility for the output and how it is used is yours. Do not use the API for anything unlawful where you are.
16What is model aliasing?+
Some catalog ids are stable labels that can point at an updated upstream model over time, so a listed id keeps working even when the version behind it moves. Re-check the model list and changelog after a sync if exact behaviour matters.