Known internally as Project Strawberry, it is intended to be more focused on reasoning, taking time to ‘think’ before it responds. It will therefore be much slower, but also more effective than its predecessors when it comes to complex tasks.

According to The Information, OpenAI could launch a new AI language model code-named ‘Strawberry’ before the end of September 2024.

Compared to previous models, this one would stand out for its ability to take its time, favouring reasoning over speed. It could thus take up to 20 seconds or so before responding in a more precise way and with reduced risk of error.

Strawberry is a text-only model. However, it is said to be capable of solving complex mathematical problems and completing advanced programming tasks. Note that this model is not the long-awaited GPT-5, which could be unveiled this winter.

Again, according to The Information, Strawberry will initially be available for paying subscribers to ChatGPT Plus.

The latest model launched by OpenAI is GPT-4o mini, which replaced the aging GPT-3.5 this summer. As a result, users of the free version of ChatGPT benefit from a more powerful tool, with no limits on usage.

Some 200 million people now use ChatGPT every week.

OpenAI is also working on SearchGPT, a new type of search engine still at the prototype stage. Like other recently launched alternatives, it aims to rival the web search leader, Google, by leveraging the possibilities offered by artificial intelligence.