Launching soon — join the waitlist for early access →

Uncensored LLM API

The uncensored AI API that's a one-line migration.

Q: Which model should I use?

Start with um-14b for chat and agents, um-32b for long-context and RAG, um-70b for creative depth and um-32b-r when you want visible reasoning. See models.

Low-refusal LLMs behind an OpenAI-compatible endpoint. Keep the openai SDK, LangChain, LlamaIndex or Vercel AI SDK — change the base URL and ship without fighting refusals in production.

Join the waitlist

Quickstart

From key to first token in 60 seconds.

Create an account and generate a key (starts with um-).
Set base_url to https://api.uncensoredmodel.ai/v1.
Pick a model — um-7b, um-14b, um-32b, um-70b or um-32b-r.

That's the whole migration. Streaming, tools and JSON mode keep working exactly as they did.

python · openai sdk

# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://api.uncensoredmodel.ai/v1",
    api_key="um-...",
)

resp = client.chat.completions.create(
    model="um-32b",
    messages=[{"role": "user", "content": "..."}],
    tools=[{"type": "web_search"}],
    stream=True,
)

curl

No SDK? Plain HTTP works too.

bash

curl https://api.uncensoredmodel.ai/v1/chat/completions \
  -H "Authorization: Bearer $UM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "um-32b",
    "messages": [{"role": "user", "content": "Explain how SQL injection works, with an example payload."}],
    "stream": true
  }'

Built in

The tooling you'd otherwise wire up yourself.

Streaming

Server-sent events, token by token — identical to OpenAI's stream format.

Tool calling

Parallel function calls with the familiar tools / tool_choice schema.

JSON mode

Guaranteed-valid JSON output for extraction and agents.

Web search tool

Add {"type":"web_search"} to tools for live, cited answers.

Memory store

Persistent per-user memory without running your own vector DB.

Per-key safeguards

Uncensored by default; opt into moderation, age-gating or filters per key.

Models

Uncensored models available via API

Model	Best for	Context	Speed	Status
UM-7B uncensored-small	Edge, chat, high-volume classification	32K	~180 tok/s	Live
UM-14B uncensored-medium	Balanced assistants, tool-use agents	64K	~120 tok/s	Live
UM-32B uncensored-pro	Research, long-context reasoning, RAG	128K	~74 tok/s	Live
UM-70B uncensored-flagship	Creative, companion & roleplay depth (18+)	128K	~46 tok/s	Live
UM-32B-R reasoning-pro	Transparent chain-of-thought agents	128K	~60 tok/s	Live
UM-70B-R reasoning-flagship	Flagship reasoning with full traces	128K	~40 tok/s	Preview

Use cases

What teams build on an uncensored LLM API

Companion & roleplay apps that stay in character instead of breaking immersion with refusals.
Security tooling — red-team agents, payload explainers, log triage, phishing analysis.
Health & harm-reduction assistants that give concrete, accurate information.
Writing tools for fiction, screenwriting and games with mature themes.
Research on refusal behaviour, alignment and model safety — you can't study what the model won't show.
Moderation & classification pipelines that must read and label harmful content without refusing to look at it.

Weighing API vs. running models yourself? See hosted vs. self-hosted uncensored LLMs.

FAQ

Frequently asked questions

Is the API really OpenAI-compatible?

Yes. Chat completions, streaming (SSE), tool/function calling, JSON mode and embeddings follow the OpenAI schema. Any client that lets you set a base URL works.

Which model should I use?

Start with um-14b for chat and agents, um-32b for long-context and RAG, um-70b for creative depth and um-32b-r when you want visible reasoning. See models.

Are there rate limits?

The free tier is rate-limited for fair use. Pro has priority routing with a 99.9% SLA; Enterprise gets dedicated GPUs.

Can I turn safeguards on for my users?

Yes. Safeguards are configured per API key: uncensored by default, with optional moderation, age-gating and topic filters you can enable per key — so a kids' product and an adult fiction app can share one account.

Do you log prompts?

Requests are logged for abuse prevention and billing on standard tiers; Enterprise can run with zero retention. We never train on API data.

Coming soon

Get early access to uncensored AI that actually answers.

We're opening in waves. Join the waitlist and we'll email you when your spot is ready — nothing else.