Salesforce CEO Marc Benioff, one of Silicon Valley’s most vocal boosters of ChatGPT, announced Sunday that he’s abandoning OpenAI’s chatbot after a brief trial of Google’s newly released Gemini 3 model.

“Holy s—. I’ve used ChatGPT every day for 3 years. Just spent 2 hours on Gemini 3. I’m not going back,” Benioff wrote on X. “The leap is insane — reasoning, speed, images, video… everything is sharper and faster. It feels like the world just changed, again.”

Benioff, who has 1.1 million followers on X, had more than 3.2 million people see his post by Tuesday morning, according to the social network. His endorsement is notable not just because of his prominence—his net worth is currently around $8.5 billion, according to Forbes—but because of his company’s extensive partnerships across the AI landscape. Just last month, Salesforce announced an expanded strategic partnership with OpenAI, integrating its Agentforce 360 platform with ChatGPT and allowing enterprises to use OpenAI’s GPT-5 models within Salesforce products.

At the time, Benioff praised that collaboration. “As consumers, we already get instant recommendations or insights from ChatGPT,” he said in October. “Now enterprises can deliver that same intelligence and immediacy.”

The swift about-face from ChatGPT to Google’s Gemini underscores how quickly things can change as AI models leapfrog one another in capability on a near-monthly basis.

Google’s big debut

Google and its DeepMind division released Gemini 3 last week, describing it as the company’s “most intelligent model, which “combines all of Gemini’s capabilities together so you can bring any idea to life.” The model almost immediately topped the LMArena leaderboard, a crowdsourced benchmark that evaluates AI systems on reasoning, coding, writing, and factual accuracy.

Leaders across Silicon Valley took notice of Gemini’s big launch. OpenAI CEO Sam Altman congratulated Google on X, saying: “Congrats to Google on Gemini 3! Looks like a great model.” Andrej Karpathy, one of OpenAI’s co-founder and a former AI director at Tesla, said on X he had “a positive early impression” of the model, calling it “very solid daily driver potential” and “clearly a tier 1 LLM.” Stripe CEO Patrick Collison posted on X that Gemini 3 successfully built him an “interactive web page summarizing 10 breakthroughs in genetics,” which he called “pretty cool.”

But behind the public compliments, there are signs of concern at OpenAI. In an internal memo written before Gemini 3’s release and obtained by The Information, Altman told employees to expect “rough vibes,” adding “by all accounts, Google has been doing excellent work recently.” He said Google’s progress could “create some temporary economic headwinds for our company,” but insisted OpenAI is “catching up fast.”

There’s been a flurry of new models in the AI sector recently. OpenAI released GPT-5.1 less than a week before Google debuted Gemini 3, and Anthropic just launched its Claude Opus 4.5 model on Monday.