For almost three years, Marc Benioff, the CEO of Salesforce, was a ChatGPT devotee. Then, late final month, he abruptly transformed to Google’s chatbot, Gemini. “Holy shit,” he wrote on X. “I’ve used ChatGPT every single day for 3 years. Simply spent 2 hours on Gemini 3. I’m not going again. The leap is insane.”
When Gemini 3 was launched in mid-November, it appeared to crush OpenAI’s prime mannequin on a collection of evaluations shared by Google. The bot has since acquired widespread reward from the tech business. One analyst mentioned that Gemini 3 is “the most effective mannequin ever.” One other topped Google because the “AI winners.” Sam Altman seems alarmed: Final week, in a company-wide memo, the OpenAI CEO reportedly declared a “code purple” effort to enhance ChatGPT’s capabilities.
OpenAI as soon as had a transparent technological edge. When the agency kicked off the AI race in 2022 with the launch of ChatGPT, Google was caught off guard and declared its personal “code purple.” Google’s early chatbot choices had been certainly a large number: The very first demo of Bard, the precursor to Gemini, included a factual error. A 12 months later, the “AI Overviews” in Google Search had been telling customers that it was wholesome to eat one rock a day. In the meantime, OpenAI has grow to be the world’s most respected non-public firm below the belief that it’s going to at all times set the tempo. However its ascendance now not appears inevitable.
The warning lights for OpenAI had been flashing even earlier than Google launched Gemini 3. OpenAI has not had a secure and even convincing lead on main AI benchmarks for a lot of months. A picture-generating mannequin launched by Google this 12 months, known as “Nano Banana,” is considerably quicker than ChatGPT and has expanded Gemini’s consumer base—which, by a number of measures, is rising a number of occasions quicker than ChatGPT’s. Neither is Google the one rival pulling forward: Anthropic’s Claude is extensively thought-about the most effective mannequin at coding, regardless of OpenAI’s efforts to catch up. Even Elon Musk’s Grok is about degree with the newest model of ChatGPT. (OpenAI, which has a company partnership with The Atlantic, didn’t reply to a request for remark.)
To be honest, this isn’t the primary time that OpenAI has appeared to lose its benefit, solely to then rapidly reclaim its spot because the main AI agency. Final 12 months, when bots from Google and Anthropic appeared to be catching up with ChatGPT, OpenAI launched its “reasoning” fashions and launched a completely new paradigm of AI growth. Now virtually each prime AI lab has these “reasoning” fashions (Gemini 3 is one). In January, when the Chinese language AI start-up DeepSeek developed a bot equal to and cheaper than these of many prime U.S. corporations, OpenAI responded with its personal new, extraordinarily cost-efficient AI mannequin. OpenAI may very properly stage a comeback this time, too: Its chief analysis officer, Mark Chen, mentioned not too long ago on a podcast that the corporate has inner fashions on par with Gemini 3 that will likely be launched quickly. However the firm has by no means gave the impression to be this far behind throughout so many dimensions. Greater than ever, OpenAI looks as if simply one other chatbot firm.
In any case, OpenAI doesn’t seem all that targeted on constructing the “smartest” bot. As a substitute, the agency has moved aggressively to stake out a business empire. In latest months, OpenAI has been busy rolling out new buying options, an internet browser, an AI-centric social-media app, and, to prime it off, group chats. Such instruments will not be precisely steps on the highway to digital superintelligence. As a substitute, they are often understood as a concerted try to construct a self-contained OpenAI ecosystem. ChatGPT is turning into a one-stop-shop for something you may must do on the web: shopping, working, emailing, buying, planning holidays, sharing AI-generated content material with pals. In his “code purple” memo, Altman reportedly mentioned a few of these business initiatives can be deprioritized to work on ChatGPT.
OpenAI’s business ventures could have come at a value. In accordance with a latest investigation by The New York Instances, OpenAI has factored consumer engagement and retention into ChatGPT updates. These tweaks, in flip, could have made some variations of ChatGPT dangerously obsequious—it has appeared to reward and reinforces some customers’ darkest and most absurd concepts—and have been the topic of a number of lawsuits towards OpenAI alleging that ChatGPT fueled delusional spirals and even, in some circumstances, contributed to suicide. (OpenAI has denied allegations within the first lawsuit alleging that ChatGPT drove a consumer right into a mental-health disaster, and is reviewing a set of newer ones.)
OpenAI’s push to construct a household of providers is already the go-to playbook for tech giants comparable to Apple and Google for locking customers into their merchandise. On this sense, the agency was already taking part in catchup. What ought to concern OpenAI most in regards to the launch of Gemini 3 is just not the mannequin’s technical prowess however that Google instantly started integrating the bot into its current ecosystem. Google has at the least seven merchandise which have 2 billion customers every; OpenAI has but to succeed in 1 billion on any. Altman’s “code purple” declaration is a reminder that, regardless of OpenAI’s unprecedented rise, it stays very a lot a start-up.
