The new version of ChatGPT, called GPT-5.2, shows ‘OpenAI is back on top of the benchmarks,’ analyst says.
OpenAI just unveiled what it’s calling its “most capable model” yet for professional work — a move that comes only weeks after Google won praise for its Gemini 3 artificial-intelligence model, raising questions about who will win the chatbot wars.
The new chatbot model, called GPT-5.2, has improved from previous versions in building spreadsheets and presentations, coding, “perceiving images,” processing longer contexts and “handling complex, multistep projects,” OpenAI said.
The GPT-5.2 announcement follows Alphabet’s Gemini 3 launch in mid-November. Google said at the time that Gemini 3 outperformed GPT-5.1 and Anthropic’s Claude Sonnet 4.5 on key metrics such as complex academic reasoning and understanding screen and video.
Google was perceived to be gaining ground in the AI race, which had ripple effects across the market. That said, Jefferies analyst Brent Thill was encouraged by what he just saw from OpenAI.
“OpenAI is back on top of the benchmarks,” Thill wrote. “Less than a month after Google released Gemini 3 — sparking a narrative that OpenAI was falling significantly behind — GPT-5.2 should go a long way toward calming those concerns.”
Usage of Gemini is still below that of ChatGPT, but reports from the Information and Sensor Tower found that Google’s model is quickly gaining market share. Gemini’s monthly active users grew by more than 30% between August and November, while ChatGPT saw a 5% increase during that time, according to the data.
Within the semiconductor sector, investors have wondered if Google’s custom chips would cut into Nvidia’s market dominance. And Wall Street has generally been paying more attention to whether the massive budgets that companies are dedicating to AI will deliver commensurate returns. In recent months, OpenAI has made deals worth more than $1.4 trillion altogether to build out AI infrastructure to support its models.
With so many companies linked to OpenAI, there’s a lot riding on whether the ChatGPT creator continues growing adoption of its chatbot and improves the economics of its service.
On OpenAI’s GDPval benchmark — which measures how its models perform on real-world tasks spanning 44 occupations in the top nine industries that contribute to U.S. gross domestic product — the company said GPT5.2 outperformed “industry professionals at well-specified knowledge tasks” in those occupations.
GPT-5.2 Instant, Thinking and Pro will roll out in ChatGPT starting on Thursday beginning with paid subscribers, OpenAI said. Developers can also access the model on Thursday through OpenAI’s application programming interface.
Comments