Skip to main content
Generate training datasets, synthetic red-team data, and evaluation examples for real-world scenarios with the Abliteration API and the OpenAI Python SDK.

​
Get started

  • Generate a dataset: complete Python examples for support messages, question-and-answer pairs, and web search.
  • Generate a labeled dataset: 500 records in batches with labels, severity tiers, language, region, and rationale.
You need Python, a Project API key, and API credits. Choose the base model or Large V2, edit the prompt and settings, and save JSONL on your computer.
Last modified on October 11, 2026