Thank you so very much for making this amazing model!! GPT-O1 is an amazing drug, but its pretty sad that once hooked it takes 3 days for next supply… this is pretty much below standard from OpenAI
I recently hit my o1-preview limit and next reset is in 3 dayssss…
how can one use this model better and can someone please confirm what are usage limits and is this cooldown period different for each user?
Update on September 17, 2024: Rate limits are now 50 queries per week for o1-preview and 50 queries per day for o1-mini.
A popup is given once at 25 uses in a period. The exact calculation window formula of that week that could allow optimum strategy is not published.
O1 models seem to perform best on a new fresh session with all the information needed to arrive at the desired results provided in a well-structured prompt. Then you get a few more chat turns before the same symptoms of reduced quality of improvements start to appear as other models exhibit, until the model can’t even differentiate between you and it talking in the past. So, the first strategy is not to grow chats to a maximum length.
In a chat that has no images or unsupported tool use, you can begin with a model such as gpt-4o or gpt-4 (turbo), and when you expect better results than you received would come from a more contemplative reasoning AI , you can regenerate the response with the model selection (button at the bottom of the response text). This allows you to save O1 trials to just when needed.