Benioff dumps ChatGPT for Google's Gemini 3, calling shift 'insane'

Marc Benioff spent two hours with Google's new Gemini 3 model on a Sunday and decided three years of daily ChatGPT use were over. The Salesforce CEO posted his verdict on X with the directness of someone who had just witnessed something he did not expect: the reasoning was sharper, the speed was faster, the images and video processing were noticeably better. "I'm not going back," he wrote. "The leap is insane."

The post landed with force. By early Monday morning, it had accumulated more than a million views—a significant endorsement from one of the most visible executives in enterprise software, and a public signal that the competitive ground in artificial intelligence had shifted again. Benioff's switch mattered partly because of who he was and partly because of what he represented: a longtime, daily user of the market leader, now convinced enough by a competitor's product to abandon it entirely after a two-hour trial.

Google had unveiled Gemini 3 the previous week through its DeepMind division, describing it as their most powerful model to date—one capable of handling text, images, video, and code with tighter integration across Google's ecosystem. The company positioned it as a step forward in reasoning and what they called "agentic" capabilities, the ability to understand and act on complex instructions. Sundar Pichai, Google's CEO, called it the company's most intelligent model, one that combined all of Gemini's capabilities into a single system.

Benioff's endorsement arrived in a landscape already crowded with competing claims. OpenAI had released ChatGPT 4.5 Turbo and was preparing to launch version 5. Anthropic had Claude 3.5 in the market. Each new release pushed against the boundaries of what these systems could reason through and accomplish. The speed of iteration had become dizzying—what seemed like a clear leader one month could be overtaken the next.

What made Benioff's statement notable was not just that he had switched, but that other prominent voices in technology quickly validated his assessment. Elon Musk, who had co-founded OpenAI before leaving its board, praised Pichai following the launch. Andrej Karpathy, who had led AI development at Tesla, posted that Gemini 3 gave him a "positive early impression" and had "very solid daily driver potential"—the kind of understated endorsement from a serious technologist that carries weight. Patrick Collison, CEO of Stripe, highlighted a specific capability: Gemini 3 had built an interactive web page summarizing ten breakthroughs in genetics, which he found impressive.

Sam Altman, OpenAI's CEO and the person with the most direct stake in the outcome, responded with grace. He congratulated Google on the launch, posting that Gemini 3 looked like "a great model." In an internal memo, Altman acknowledged what he called "temporary economic headwinds" but expressed confidence in OpenAI's ability to weather competition and maintain its leadership position. He emphasized that the company had built enough strength to absorb the arrival of capable models elsewhere, and that the focus now needed to be on reaching superintelligence—the next frontier entirely.

Benioff's position carried particular weight because Salesforce maintained partnerships across the AI ecosystem, including with both OpenAI and Anthropic. He was not a partisan voice; he was a pragmatist watching the technology evolve and making a choice based on what he observed. His willingness to say so publicly, and to say it in terms that suggested a genuine shift rather than a marginal improvement, signaled that the competitive dynamics in AI had entered a new phase. The question was no longer whether alternatives to ChatGPT could match its capabilities. The question was whether they had begun to exceed them.