Why is ChatGPT limiting free users to GPT-5.2 Instant?

OpenAI is defaulting free users to GPT-5.2 Instant as a cost-saving measure, though they state it's also an improvement for user choice.

Can I still use the advanced 'Thinking' model on ChatGPT for free?

Yes, free users can still access the 'Thinking' model, but they must now select it manually from the tools menu.

What is the difference between GPT-5.2 Instant and Thinking models?

GPT-5.2 Instant is described as a powerful workhorse for everyday tasks, while the Thinking model solves harder work tasks more effectively.

Home / Technology / ChatGPT Dumbs Down for Free Users

ChatGPT Dumbs Down for Free Users

17 Dec

•

Summary

Free ChatGPT users now default to GPT-5.2 Instant model.
Paid users and those in specific regions have access to higher tiers.
Change is a cost-saving measure for OpenAI, potentially impacting user experience.

OpenAI has implemented a significant change for its free ChatGPT users, defaulting their interactions to the GPT-5.2 Instant model. This update, noted on December 11, means that queries from free users and those on the ChatGPT Go plan will no longer automatically be routed to more advanced AI models. Previously, ChatGPT would switch to a more capable 'Thinking' model for complex queries.

While free and Go subscribers can still manually access the 'Thinking' model via the tools menu, this requires user intervention. OpenAI describes the Instant model as suitable for everyday tasks, contrasting it with the 'Thinking' model's effectiveness for harder work. This shift is viewed as a cost-saving strategy for OpenAI, potentially impacting the quality of responses for a large user base.

The company suggests that GPT-5.2 Instant is now better equipped for sensitive queries, a role previously handled by the reasoning model. This decision follows user frustrations with earlier versions of ChatGPT and internal acknowledgments from OpenAI about user dissatisfaction with model selection. The long-term impact on user experience and the quality of AI assistance remains to be seen.