Maker of ChatGPT OpenAI is developing a new method for its AI models under a project code-named “Strawberry,” as per internal documents seen by Reuters and a person with knowledge of the situation.

The project—details of which have not yet been made public—occurs as the Microsoft-backed business rushes to demonstrate that the models it provides can provide sophisticated reasoning powers.

According to a copy of a recent internal OpenAI document reviewed by Reuters in May, teams within OpenAI are working on Strawberry. The document, which describes a blueprint for how OpenAI plans to use Strawberry to do research, was not available for Reuters to verify the exact date. The proposal is still a work in progress, the source told Reuters. The news organization was unable to determine Strawberry’s distance from public availability.

Even inside OpenAI, the workings of Strawberry are well guarded secrets, according to the source.

A representative with OpenAI responded in a statement when questioned about Strawberry and the information in this story, saying, “We want our AI models to see and understand the world more like we do.” In the industry, it is customary to conduct ongoing research into new AI capabilities with the shared premise that these systems will eventually become more intelligent.

Questions concerning Strawberry were not explicitly addressed by the representative.

Formerly branded as Q*, the Strawberry project was already viewed as a breakthrough within the corporation, according to a Reuters report from last year.

Two people reported seeing what OpenAI employees claimed to be Q* demos earlier this year, which are able to provide answers to challenging math and science problems beyond the capabilities of currently existing commercial models.