OpenAI is experimenting with a weekly limit of 3,000 for GPT-5 usage.

OpenAI has addressed criticism regarding the token limits of GPT-5 by introducing a new limit of 3,000 Thinking messages per week for Plus users, as confirmed by Sam Altman in a series of posts on X. This change aims to enhance the reasoning capabilities of the model while also indicating that OpenAI plans to raise all model-class rate limits above pre-GPT-5 levels, potentially making GPT-5 one of the most cost-effective options available. Users had previously expressed confusion over when GPT-5 switches between its Thinking and smaller models, like GPT-5-mini. To alleviate this issue, OpenAI is testing a new user interface indicator to clarify which model is in use. Altman also noted a significant increase in the daily usage of reasoning models, as GPT-5 can automatically route queries to different models. 

Categories: 1. **Token Limits and Pricing Strategy**
2. **User Experience and Model Switching**
3. **Model Utilization and Performance** 

Tags: 1. OpenAI
2. GPT-5
3. Token limits
4. Cost
5. Profit
6. 3,000-per-week limit
7. Plus users
8. Reasoning models
9. UI indicator
10. Model-class rate limits