According to reporting by The Information, OpenAI might be releasing its Project Strawberry by the end of this month within ChatGPT. The report from The Information is based on details from two testers who were apparently involved with the model. Project Strawberry, previously known as Q*, is supposed to be a reasoning model that will give ChatGPT the ability to think before it responds.

Hence, it will make the existing models more accurate but it will also lead to ChatGPT taking longer to respond. The model will reportedly be available from the drop-down menu within ChatGPT so users can decide which model to use.

What do we know about the Strawberry model? The model is supposed to be using chain-of-thought promoting enabling it to perform multi-step tasks based on a single query. It can even plan ahead and navigate the internet autonomously to perform "deep research" for complex queries. While the time taken for each response will be longer, it will also make it less likely to hallucinate.

Speculations say that it will only be released as a text-based model initially, with no multimodal support, i.e., no support for vision, images, or videos.

OpenAI's CEO Sam Altman posted a photo of his strawberry plant on X last month and people are now making the connection between the two. While it could be that people are grasping at straws here, knowing Sam Altman's previous teases on X, there could be something here.

Currently, there's no knowing whether Project Strawberry will be released to ChatGPT subscribers within the same subscription or if the company will introduce another tier, given the cost overhead that has to be involved with performing such complex tasks. Perhaps OpenAI might introduce rate limits to combat that. However, one thing seems to be sure: it is highly unlikely that ChatGPT-free users will get access to the model.

Early impressions also indicate that the model is only slightly better than GPT-4o (but it can handle longer tasks better), and its performance can be sometimes underwhelming. It can even make mistakes and also struggles with memory sometimes, as per the report. So, even if OpenAI indeed releases the model by the end of this month, it might not turn out to be anything too game-changing. But only time will tell.