Skip to content
Gods enthroned on clouds in a painted ceiling: a goddess flying overhead, Minerva with her spear, Cybele in her lion chariot, Neptune with his trident

Abliterated open models, with zero data retention.

Stop getting denied by AI.

Get started Talk to us
Now serving, across providers:

François Lemoyne, The Apotheosis of Hercules (detail), 1733–36. Versailles. Lemoyne, The Apotheosis of Hercules, 1733–36.

Use the SDK you already have

Lutheos speaks both the OpenAI and the Anthropic API. Set the base URL, your key and a model name; the rest of your code stays as it is.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.LUTHEOS_API_KEY,
  baseURL: "https://lutheos.com/v1",
});
import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  apiKey: process.env.LUTHEOS_API_KEY,
  baseURL: "https://lutheos.com",
});

Works with the tools you use

Any SDK, framework or editor that lets you set a base URL can call Lutheos.

OpenAI SDK
https://lutheos.com/v1
Anthropic SDK
https://lutheos.com The SDK adds the /v1 itself.
Such as
  • OpenAI SDK
  • Anthropic SDK
  • LangChain
  • cURL
  • Claude Code
  • Cursor
  • VS Code
  • Cline
  • Zed

The request keeps its shape

Streaming, function calls and vision pass through unchanged.

  • Both APIs: the OpenAI and the Anthropic SDK reach the same models through one endpoint.
  • One key: every request draws from one balance, at its model's listed price.
The full quickstart

Pricing

Pay per token, from a balance you top up. No monthly fee.

Get started See pricing

Prices

Prices per 1M tokens, in US dollars
ModelInputOutputCached inputContextMax output
DeepSeek V4.1 Flash 1017 Abliterated Refusals removed$0.10$0.50$0.021M384K

Prices are per 1M tokens. Cached input is the price of input tokens served from the provider's cache. Context and max output are in tokens.

Billing

A prepaid balance
Top up a balance and pay for the tokens you use. Nothing renews: no subscription, no seats, no monthly fee.
Input and output
Counted separately, each at the model's listed price per million tokens.
Failed requests are free
Requests that fail upstream are not billed.
A key for every project
Give each project its own key, and revoke one at any time. Every key draws from the same balance.
Pay in crypto
159 coins, including BNB, BTC and ETH. Top up from the dashboard.

Guido Reni, Aurora (detail), c. 1612–14. Rome.

Questions

Something missing? Write to [email protected]

Read the docs

  • A private AI inference API for the open models our GPU partner self-hosts with zero data retention, abliterated and as released. One endpoint speaks both the OpenAI and the Anthropic API.

  • The ones our partner hosts. An ID ending in -abliterated is a model changed to refuse far less; one ending in -non-abliterated is the model as released. Read what abliterated means, or see the live list and prices on the models page.

  • The partner that runs these models does not keep your prompts or the replies once a request is done.

  • No. It works with the OpenAI and Anthropic SDKs: change the base URL and the model name. The docs show both.

  • Pay as you go from one balance, at each model's listed price per million tokens. There are no monthly fees, and upstream errors are not billed. Each project can have its own key, and all keys draw from that balance. See pricing.

  • Sign in, create a key, and top up your balance in crypto from the dashboard.

Every model we serve, on one key.

One endpoint for the OpenAI and Anthropic SDKs, on GPUs that keep no data. Top up a balance in crypto and pay for the tokens you use.

Get started Read the docs

François Lemoyne, The Apotheosis of Hercules (detail), 1733–36. Versailles. Lemoyne, The Apotheosis of Hercules, 1733–36.