Starting in December, ChatGPT Plus, Pro, and Business subscriptions will share a single usage pool for chatting and coding with Codex.
The way usage is measured is changing too. It used to be counted by messages; now it’s counted by tokens. The more advanced the model and the more complex the task, the more tokens it uses. One long reasoning session could use as much as more than ten messages used to.
There’s just one exception: Instant chats don’t count toward your limit. OpenAI says you can use Instant as much as you like under your current plan.
So how big is the pool in total? No one knows. The help documentation doesn’t give a specific number. No one knows how many tokens you’ll get in December. OpenAI has only said that it expects most users won’t have to wait for a reset more often than they do now.
If you only chat and don’t use Codex, the half of your allowance that would otherwise sit idle gets added to the shared pool, so you might actually have more room. The same goes in reverse if you only use Codex to code. The tightest spot is for people who use both heavily: both sides will be drawing from the same pool.
Looking back at OpenAI this October, Codex limits were reset 11 times. At the end of the month, on Day 25, OpenAI announced that usage limits across all Pro and Business plans would double, and the whole internet cheered. Two days later, a line of fine print appeared on the checkout page: starting in December, plan usage would share the same limit across all ChatGPT and Codex features. Plus’s biggest selling point was that you didn’t have to keep an eye on your usage limits. After the pools are combined, that advantage is gone.
OpenAI is bringing Work’s agent capabilities into ChatGPT’s main interface, so users can run tasks directly in the chat window. An agent task can use as many tokens as an average person’s chats over several days—usage like that can’t be priced by the message. Combining the pools is a side effect of this shift, and other companies building agents will eventually do the same.
If you’re a heavy user, move the demanding tasks to Instant first—it’s outside the pool.
#OpenAI #ChatGPT #Codex
The way usage is measured is changing too. It used to be counted by messages; now it’s counted by tokens. The more advanced the model and the more complex the task, the more tokens it uses. One long reasoning session could use as much as more than ten messages used to.
There’s just one exception: Instant chats don’t count toward your limit. OpenAI says you can use Instant as much as you like under your current plan.
So how big is the pool in total? No one knows. The help documentation doesn’t give a specific number. No one knows how many tokens you’ll get in December. OpenAI has only said that it expects most users won’t have to wait for a reset more often than they do now.
If you only chat and don’t use Codex, the half of your allowance that would otherwise sit idle gets added to the shared pool, so you might actually have more room. The same goes in reverse if you only use Codex to code. The tightest spot is for people who use both heavily: both sides will be drawing from the same pool.
Looking back at OpenAI this October, Codex limits were reset 11 times. At the end of the month, on Day 25, OpenAI announced that usage limits across all Pro and Business plans would double, and the whole internet cheered. Two days later, a line of fine print appeared on the checkout page: starting in December, plan usage would share the same limit across all ChatGPT and Codex features. Plus’s biggest selling point was that you didn’t have to keep an eye on your usage limits. After the pools are combined, that advantage is gone.
OpenAI is bringing Work’s agent capabilities into ChatGPT’s main interface, so users can run tasks directly in the chat window. An agent task can use as many tokens as an average person’s chats over several days—usage like that can’t be priced by the message. Combining the pools is a side effect of this shift, and other companies building agents will eventually do the same.
If you’re a heavy user, move the demanding tasks to Instant first—it’s outside the pool.
#OpenAI #ChatGPT #Codex