AI: The frantic race for performance

For several months now, tech giants have been developing and improving their artificial intelligence assistants. This race for performance is becoming increasingly frenetic, creating a tacit competition within companies. OpenAI, a month after launching GPT-5.1, unveiled a new version of ChatGPT, GPT-5.2, at the beginning of December, in response to the new version of the Google AI assistant, Gemini 3, revealed at the end of November this year.

A smartphone displays the Google Gemini 3 interface with on-screen options such as “Create image,” “Write anything,” and “Help me learn,” while the Gemini 3 logo appears on a larger screen in the background in San Ferdinando di Puglia, Italy, on December 4, 2025. The image highlights the user interface and branding of Google's AI platform (Photo by Matteo Della Torre/NurPhoto via Getty Images)

These two giants have been battling it out for the title of AI assistant champion for several months now. This rivalry between Google and OpenAI has intensified since the respective unveiling of their new AI assistant versions: Gemini 3 for Google and ChatGPT-5.2 for OpenAI.

Data from Sensor Tower, which analyzes the behavior of 5 million consumers worldwide, shows rapid growth for Gemini since the launch of Gemini 3 at the end of November. The number of worldwide visits to the Gemini site is said to have doubled between August and November, while those to ChatGPT have only increased by around 1%. Over the same period, Gemini would have reached 346 million monthly active users, an increase of close to 30%, while ChatGPT peaked at 810 million, with an increase limited to just under 5%.

This discrepancy in dynamics explains Sam Altman's decision to launch an internal “Code Red” operation to refocus teams on ChatGPT. This internal climate enabled the OpenAI teams to have the resources they needed to finalize the new version of the AI assistant. The unveiling of the new iteration of ChatGPT last December saw the birth of a version of GPT that is more efficient than ever, according to the company, reducing errors relating to the“thinking” version of the assistant by 38%.

ANKARA, TURKEY – DECEMBER 30: Logos of ChatGPT and OpenAI are displayed on cell phone screens in Ankara, Turkey on December 30, 2025. (Photo by Omer Taha Cetin/Anadolu via Getty Images)

Despite GPT's developments and their 800 million operators every week, OpenAI has declared that it is losing money every month and only expects to be profitable in 2029. An entirely different scenario is being experienced on the Google side, where the company is generating billions of dollars every quarter from its other activities.

At this stage, the conversational robot competition is essentially based on two players, even if other competitors are well and truly in the game. Claude, Copilot, Deepseek, etc. are all key protagonists in this frantic race to improve artificial intelligence tools. A fundamental question remains: how far can artificial intelligence evolve without redefining the place of humans, and above all, will this frantic race for performance ever come to an end?