3,000-Message Cap: OpenAI Trials Usage Limits for GPT-5’s Advanced Reasoning

Since the introduction of GPT-5, criticism has been hailing: Users have been bothered by the limits and non-transparency when the system switches to smaller models. Openai is now trying to smooth the waves – and increase acceptance for the model.

Openai tests weekly limit for GPT-5 Thinking

Openaai plans to regulate the use of its current flagship model GPT-5 and to communicate it very clearly. For the extended “thinking” function, which enables particularly complex logical conclusions and complex analyzes, the company is currently testing a limit of 3,000 messages per week for plus subscribers (via Bleeping computer). That hadn’t been said so clearly. According to Openai boss Sam Altman, this limit should not only raise the previous usage limits, but also create a balance between costs, performance and demand.

In a number of articles on the X platform, Altman announced that long -term all model classes – including smaller variants – should be unlocked beyond the stand before the introduction of GPT -5. One goal is to make GPT-5 one of the most cost-effective and most powerful models on the market. In addition to the limit adjustments, Openai is working on a function that should clearly display in the future when GPT-5 thinking will be active and when automatically switches to a smaller variant such as GPT-5-mini.

This transparency is said to solve a frequently expressed user problem: so far it has not always been seen which model was answering. According to Altman, the use of the more complex “Reasoning” models has increased significantly in recent months. While they used to be used by one percent of free users, the proportion is now around seven percent.

Overview Problems with the GPT-5 start:

• Openaai experienced massive user criticism after the start of GPT-5, especially because of changed limits and the shutdown of older models such as GPT-4O, which caused displeasure to pay customers.
• CEO Sam Altman admitted technical problems with the rollout and promised corrections, including the restoration of access to GPT-4O for plus subscribers.

In the case of plus subscribers, the proportion grew from seven to 24 percent – and the trend increases. How Openaai wants to divide the server capacities between different user groups and types of application in the coming months should be announced shortly. Comparisons between Chatgpt and the API, between existing and new customers or between research and product applications could play a role.