LLM models have a token limit. Everything you type counts toward this limit.
The “xx% context left” message refers to how much of the available token context window is remaining when you’re interacting with Codex.
doggy8088· 1 year ago
I understand now. Thanks.
I chose o4-mini, which has 200,000 tokens. The more history records there are, the more tokens are consumed.
batchu· 1 year ago
Is this limit daily or monthly?
Fontanjo· 1 year ago
Is this limit daily or monthly?
This is not a subscription limit. It has to do with the model capacity. The context is what the model can "see". Transform-based LLM don´t really have a "memory", they just read all the past conversation as input every time. For this reason, the context left will reset to 100% for every new chat, and will slowly fill during the conversation.
MarshallZzzz· 11 months ago
Is there a way to reset this? Possibly deleting the past inputs
6 Comments
LLM models have a token limit. Everything you type counts toward this limit.
The “xx% context left” message refers to how much of the available token context window is remaining when you’re interacting with Codex.
I understand now. Thanks.
I chose
o4-mini, which has200,000tokens. The more history records there are, the more tokens are consumed.Is this limit daily or monthly?
This is not a subscription limit. It has to do with the model capacity. The context is what the model can "see". Transform-based LLM don´t really have a "memory", they just read all the past conversation as input every time. For this reason, the context left will reset to 100% for every new chat, and will slowly fill during the conversation.
Is there a way to reset this? Possibly deleting the past inputs
/compact