I’ve noticed that the current token limit per message for the recently released o3 and o4-mini models is around 60-65K tokens in the chat app, which feels quite restrictive compared to previous models (now retired) that allowed up to approximately 100K tokens. Considering the theoretical limit is around 200K tokens, I’m wondering if this limitation is intentional or if it’s a bug.
I’m currently subscribed to ChatGPT Pro specifically for professional usage, and hitting this lower ceiling makes the tool almost unusable for my purposes. Even more frustrating, the macOS app has an even stricter limit, not allowing me to paste more than around 30K tokens per message.
Could someone from OpenAI clarify if these limits are permanent? If they are, I’ll have to reconsider the value of maintaining a Pro subscription, as these constraints seriously affect productivity.
This is serious problem for me as well. It means you have to split the code into more posts. Which I have found reduces the models correctness. It will also use more of the 50 responses per week limit. I honestly think I will have to try out other AI vendors if this cant be resolved. Is there a reason we cant just have a single post that can handle around 150k - 175k tokens? This would leave enough context left for output responses.
I’m experiencing the exact same issue. I’m also on the Pro plan, and right now it feels like even a 400-line code snippet gets cut off, which completely breaks the workflow.
What’s even more frustrating is that when trying to preview code in the interface, it won’t render more than ~200 lines—something previous models like o3-mini-high handled with no problem. This makes it really hard to use for any serious development work.
Considering this is a paid product that used to offer much more generous input and output lengths, these limitations feel like a step backward. If this is going to be the new standard, it seriously impacts the value of the Pro subscription.