Does anybody know what the token-limit of the new version GPT 4o is? Unfortunately, during my research on this topic, I keep finding different pieces of information.
3 Likes
128k tokens
https://platform.openai.com/docs/models/gpt-4o
thats for API
chatGPT is probably 32k, however it seems they removed the context from chatgpt pricing page
https://openai.com/chatgpt/pricing/
so Iam not sure if its still 32k in chatGPT
EDIT: ah wait, just scroll down more 
Also from what I have seen, the tokenizer is now completely different so something like this will no longer show it correctly for gpt4o
https://platform.openai.com/tokenizer
oh actually it now says “gpt4o coming soon”
1 Like
GPT-4o, like other recent models, will not allow you to produce more than 4k of output tokens however, and is trained to curtail its responses even more than that.
1 Like
Does it the limit reset after some time? I heard that it resets every 24h
Welcome to the Forum!
Are you referring to rate limits, such as RPM (requests per minute), RPD (requests per day), TPM (tokens per minute), TPD (tokens per day)?
1 Like
Once you have paid the token amount to use the API, there are no daily limits. Just per-minute limits that far exceed that needed for several people. The gpt-4o model has the highest limits yet, letting you know how little computational impact the model has (the quality in anything but chatting also lets you know how little computation powers it…)
| Model | Tier 1 Token limits | Tier 1 Request limits |
|---|---|---|
| gpt-4o | 30,000 TPM | 500 RPM |
…
| Model | Tier 5 Token limits | Tier 5 Request limits |
|---|---|---|
| gpt-4o | 5,000,000 TPM | 10,000 RPM |
The first tier, paying $5, the rate is sufficient to empty your balance in 20 minutes if you ran a script.
The actual way the rate limit algorithm works can be described like:
- You start with a minute’s worth of requests;
- Your bucket refills at the continuous rate after use.
Although you can also get limited if you try to make all 500 RPM the same second.
I asking in regards to context lenght (not API) but browser based using GPT 4o on: chatgpt com
1 Like
My response was to a different post in your thread 
Unfortunately, I am also not fully certain about the context length for the new model on the ChatGPT interface.
I disagree that the 500 tokens per minute “far exceed what’s needed for several people”, I just made a translator app for myself and if the person speaks for a long time, it cuts it off, and we’re talking like 1:30 or 2 mins of speech, nothing unrealistic. Whisper aces it, writes it down, sends it for translation and the translation comes back truncated if it goes over the 500 tokens per minute.
1 Like
That’s 500 REQUESTS per minute. Over 8 per second.
If you are getting truncation, you should check the max_tokens API parameters and log any error messages you receive - the rate limit is not subtle, you get denied further API calls.
you’re absolutely right, that was it, my bad
1 Like
I also looked up information to find out if it was advisable to stop paying, but on Reddit I found users who say they reached 7, 13, or 16 messages before it switched from GPT-4 to @ 3.5.
And asking chatGPT4o directly, he searched the internet and it says 80 every 3 hours for a payment plan. 160 for team membership
I think the variation in the free plan is due to “variations depending on availability”
I hope i’ve helped
1 Like
I had a doubt that the Plus plan shows 5x more messages. But what is 5x more messages? Like if I have 5 messages limit per day then I will have 25? Just 25? Or is it like I will have 25 per chat and no reset time? or which plan has no reset time?
Thank you in advance
i bought the subscription based version, 37.99 NZD a month. I didnt know anything about it or request limits. i just wanted to have philosophical conversations and discuss random tidbits. Do i have some kind of limit per day? i did use a lot of requests…?
As mentioned right above the last two replies, the evaluation period for rate-limiting ChatGPT is three hours.
If you were given 80 message per 3 hours, then you’d not be able to receive an answer to question 81 until question 1 had expired by being asked over three hours ago.
1 Like
So is it that there is only 3 hours limit from starting the first question?
And is it 80 chats or 80 questions in 1 chat?
Well, any question. If you asked 40 GPT-4 questions over a four-hour period, the initial 10 you might have asked that are past hour three no longer count against the limit. A continuous rate of one input per 4.5 minutes. With GPT-4o counting against the limit half as much.
Per account.
Is it for ChatGPT Plus or Team? which one has more chats
Teams is a higher tier, but requires at least two paid seats at the higher cost, and doesn’t have the required management to prevent users from just adding more users to a company.
Here’s where current and any changing limits would be: https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus
I’m currently using the GPT-4 API with a 4K token limit, as confirmed in the Playground. How can I increase the maximum token count to 128K?