Is there any public commitment to lifecycle of gpt-4o-mini and gpt-4o finetune? Is the expectation we need to re-train new finetunes soon, or is there any commitment how long they will be available? If no, any educated guess? Old models have deprecated surprisingly quickly.
I’d say it is probably difficult to put an exact timeline to it. But I’d imagine that fine-tuned models will remain available for consumption for a significant period in time.
The approach to older models is a good case in point in my view. While the fine-tuning itself is set to be deprecated for these, the already fine-tuned models remain available. I’d expect a similar stance for gpt-4o-mini and gpt-4o fine-tunes.
Thanks! I had misunderstood the notification, and I thought the finetunes would go also away if the base model is no longer running in production.