Change three lines. Get answers, not refusals.

Dolphin LLM is an uncensored open-weight model behind an OpenAI-compatible API. Point your client at our base URL, keep the rest of your code, and pay only for the tokens you use.

Get an API key Read the docs

$0.50 of free credit for 7 days. No card, no phone number.

main.py

from openai import OpenAI client = OpenAI(    base_url="https://new.dolphinllm.com/v1",    api_key="dlm_live_…",)reply = client.chat.completions.create(    model="uncensored",    messages=[{"role": "user",               "content": "End a heist: the thief wins."}],)

import OpenAI from "openai"; const client = new OpenAI({  baseURL: "https://new.dolphinllm.com/v1",  apiKey: "dlm_live_…",});const reply = await client.chat.completions.create({  model: "uncensored",  messages: [{ role: "user",              content: "End a heist: the thief wins." }],});

curl https://new.dolphinllm.com/v1/chat/completions \  -H "Authorization: Bearer dlm_live_…" \  -H "Content-Type: application/json" \  -d '{    "model": "uncensored",    "messages": [{"role": "user",      "content": "End a heist: the thief wins."}]  }'

The lines you change. Everything else stays as it is.

Replymodel: uncensored

Mara walked out through the front doors at noon, waved to the guard she had paid in March, and was three borders away before anyone opened the vault.

The numbers

One model, one price list. No monthly fee, no per-request fee, and failed or refused requests are free.

Input tokens: $0.15 per million
Output tokens: $0.50 per million
Context window: 64,000 tokensPrompt and answer together.
Speed limits: 300 requests a minute8 at once per key. Trial keys: 60 a minute, 2 at once.
Free trial: $0.50 of creditValid 7 days, one per person, no card.
Top-ups: $20–$500By card or crypto (USDT, USDC). Credit never expires. +5% from $50, +10% from $100, +15% from $200.

A typical chat turn of 4,000 tokens in and 2,000 out costs $0.0016. More examples and a cost calculator

What the API does, and what it does not

We host one chat model and do it well. If you need the items on the right, use a bigger provider.

In the API

Chat completions in the OpenAI formatPOST /v1/chat/completions and GET /v1/models
Streaming over server-sent eventsWith token usage at the end of the stream
Function calling and JSON modetools, tool_choice, response_format json_object
Your prompts are not used for trainingWe keep request counts and token totals, not your text

Not offered

Embeddings, images, audio or video
An uptime SLA or a dedicated account manager
SOC 2, HIPAA or other compliance paperwork
On-premise installs

From sign-up to first answer

About two minutes. Email or Google, no phone number.

Create an account
Sign up with an email and password or with Google.
Copy your key
The key is in your account with $0.50 of trial credit already on it.
Top up when you need more
From $20, by card or crypto. Full speed limits open after the first top-up.

What people build with it

Fiction and roleplay

Stories with villains, violence and adult themes, without a lecture after every paragraph.

Security research

Red-team prompts and attack write-ups for testing your own defences.

Work with hard material

Summaries of court files, medical notes and extremist content for moderation teams.

Agents and tools

Function calling and JSON mode for agents that should not stop halfway on a refusal.

One hard rule

Uncensored means the model does not refuse lawful adult requests. It does not mean anything goes.

Adult content is allowed. Sexual content involving minors is always refused, with an HTTP 403 and the code content_blocked.

You must be an adult to use the service, and you are responsible for following the law where you live.

Questions and answers

Is this OpenAI?

No. Dolphin LLM is an independent service that speaks the same API format, so the official OpenAI libraries work with it. The model is not GPT, Claude or Gemini.

Which model is it?

An open-weight model hosted on our own GPU servers, with the refusal behaviour removed. Its model id is "uncensored".

Do you train on my prompts?

No. Prompts and answers are not used for training. We keep counters per key: requests and tokens.

Is there a monthly fee?

No. You top up a prepaid balance and every request takes its token cost from it. When the balance reaches zero, requests stop with HTTP 402.

How does the trial work?

Every new account gets $0.50 of credit for 7 days, one per person, without a card. Trial keys run at 60 requests a minute and 2 at once.

Your first answer is a few minutes away.

$0.50 of free credit. Keep your code, change the base URL.