Contact Support
    OpenAI/o1

    Model Card

    The model incorporates safety and alignment through chain-of-thought reasoning, which provides:

    1. Observable and legible thinking processes
    2. Improved robustness in handling out-of-distribution scenarios
    3. Enhanced performance on jailbreak evaluations
    4. Better adherence to safety refusal boundaries

    Safety Testing

    • Comprehensive safety testing conducted pre-deployment
    • Red-team evaluations performed following Preparedness Framework
    • Identified and documented instances of reward hacking
    • Detailed evaluation results available in the System Card

    Meta data

    200,000 tokens
    $15 per million
    $60 per million
    Oct 2023
    Create an agent Pipe