This Friday, OpenAI released the long-awaited GPT-5 series of models, which have made significant progress in reasoning, coding, and tool calling. All users can use GPT-5 models for free through ChatGPT. However, the degree of intelligence and usage restrictions of GPT-5 are also different depending on the ChatGPT subscription level. Free users can only get a limited number of high-smart replies, while paid subscription users can get smart replies with a higher quota.

Specifically, the quotas for different subscription levels are as follows:

ChatGPT Free users can send up to 10 messages every 5 hours. Once the quota is reached, ChatGPT will automatically switch to the GPT-5 mini model. Free users can also request a GPT-5 Thinking message once a day.

ChatGPT Plus/Team subscribers can send up to 80 messages every 3 hours. After reaching the quota, they will switch to the GPT-5 mini model until they automatically switch back when the time is up. They can also manually select GPT-5 Thinking and use 200 messages per week.

ChatGPT Pro subscribers receive unlimited access to GPT-5 models.

ChatGPT Plus subscribers are dissatisfied with restrictions:

In the past, Plus subscribers had unlimited access to the o3 and o4-mini Thinking models. Now, after the release of GPT-5, they can only use 200 Thinking messages. Now, after the release of GPT-5, the usage quota has been restricted. Therefore, users who paid to subscribe to ChatGPT Plus are very dissatisfied with this update, and they also believe that the GPT-5 model does not meet the expected capabilities.

Sam Altman personally apologized and promised improvements:

In response to protests from ChatGPT Plus subscribers, OpenAI CEO Sam Altman apologized to users on X and promised improvements, including the following:

1. Adjust ChatGTP Plus subscriber usage quota and double the GPT-5 quota, which is 160 messages every 3 hours.

2. Reflection on the model ChatGPT Plus users are now protesting the return to the 4o series model rather than being forced to use GPT-5 and the existence of a quota limit of 200 messages per week.

3. As for the problem that the GPT-5 model is not as good as expected, Sam Altman explained that this is a BUG. Due to internal routing problems, GPT-5 looks clumsy. Now OpenAI has fixed the problem.