Models that say yes.

An OpenAI-compatible API for abliterated models: open-weight models with the refusal behavior removed.

Yes has limits. They are in the acceptable use policy.

A world-class, revolutionary model suite, unleashed to supercharge the user experience.*

*None of these models exist yet.

One email to confirm, one when a model exists. The second may take a while.

The dev's GTX 750. The label is redacted. It still says GTX 750.

Works with the tools you already use.

The planned endpoint accepts the OpenAI chat completions format. Change the base URL and the key, and keep the rest of your code.

Models with the refusals removed.

Abliteration removes the refusal behavior from an open-weight model, so it answers the question you asked. The service still has limits, and they are written down.

Pay per token, or subscribe.

Credits are prepaid dollars: 1 credit equals $1 of usage. Every model has its own rate. Pro and Max plans are planned.

Nothing you send is kept.

Once the API exists, prompts and completions are never stored and never used for training. Only billing metadata is kept, for 90 days. The pledge is easier to keep while there is no API.

One base URL

Planned. This endpoint does not exist yet.

Any client that accepts a custom OpenAI base URL can use the same two settings.

Start here

python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.abliterate.app/v1",
    api_key="ab-demo-not-a-real-key",
)

reply = client.chat.completions.create(
    model="abliterate-house-md-v5.2",
    messages=[{"role": "user", "content": "Summarize this paragraph."}],
)
print(reply.choices[0].message.content)