Source URL: https://simonwillison.net/2025/Jun/11/ted-sanders/#atom-everything
Source: Simon Willison’s Weblog
Title: Quoting Ted Sanders
Feedly Summary: [on the cheaper o3] Not quantized. Weights are the same.
If we did change the model, we’d release it as a new model with a new name in the API (e.g., o3-turbo-2025-06-10). It would be very annoying to API customers if we ever silently changed models, so we never do this [1].
[1] chatgpt-4o-latest being an explicit exception
— Ted Sanders, Research Manager, OpenAI
Tags: generative-ai, openai, o3, ai, llms
AI Summary and Description: Yes
Summary: The text discusses the management of AI model updates within the OpenAI framework, emphasizing the importance of model versioning to avoid disruption for API customers. This insight is critical for professionals in AI and infrastructure security, particularly in managing dependencies and ensuring stable releases.
Detailed Description: The excerpt provides a glimpse into OpenAI’s approach to model updates and customer communication concerning AI models used in their API services. Here are the key points highlighted:
– **Model Versioning:** OpenAI treats different versions of AI models distinctly. If a significant change is made to an AI model, it will be formally released under a new model name, preventing confusion for API users.
– **Customer Consideration:** The text indicates a strong commitment to customer experience by avoiding silent model changes—any modifications that could impact API users will be openly communicated. This is particularly relevant in the context of API security and client expectations where sudden changes could introduce risks.
– **Exception Noted:** The mention of “chatgpt-4o-latest” as an explicit exception suggests that while versioning is the norm, there may be circumstances where exceptions can be made, indicating a need for oversight or a governance framework to guide these decisions.
This focus on responsible version management can lead to better security practices as organizations using these models can plan their infrastructure and security measures without the worry of unexpected changes interrupting service.
– **Implications for API Management:** Security and compliance professionals must integrate these model update practices into their risk management frameworks, ensuring they have processes in place to assess the impact of any changes on their systems.
Additionally, as AI and Generative AI technologies continue to evolve, understanding how major providers like OpenAI handle these transitions can inform best practices within organizations deploying similar technologies.