Google is back on top and not quietly. Gemini 3 didn’t just close the gap with OpenAI, Anthropic, and xAI - it seems to have flipped the leaderboard. The graph above shows that for the very first time Google’s Gemini line overtakes OpenAI’s entire GPT family in the Artificial Analysis Intelligence Index, with Gemini 3 Pro breaking past the long-standing frontier held by GPT models. For exactly three years the narrative was that Google was crawling behind while others raced ahead. That story ends today. Gemini 3 Pro beats GPT-5.1 on the benchmarks that actually matter - HLE, ARC-AGI-2, long-horizon agent runs - and with serious margins.
What strikes us the most at AI World is the user sentiment. While the release of GPT-5 in summer was anticlimactic - power users on forums and social media are openly saying Gemini 3 feels more coherent in reasoning, sharper in multimodality, and just absurdly strong in zero-shot code generation. Anthropic still shines on safety and writing, and xAI has its context windows and pricing play, but Gemini 3 is outpacing both where the frontier is moving: agentic behavior, video understanding, screen parsing, tool use.
But the story no one is talking about? Hardware. Everyone’s obsessing over benchmarks, but Google’s integrated TPU stack is the bigger story - finally delivering the efficiency and scale they’ve been promising for years. OpenAI and Anthropic are renting (very, very expensive) compute. Google owns it. It seems that this comparative advantage is starting to compound over time. When you control your entire pipeline, you can ship models faster, cheaper, and larger without asking permission or draining your bank account.
So yes, the leaderboard flipped today. For November 2025, Google is the lab to beat. Will OpenAI or Anthropic counterpunch in the coming weeks or months? Probably. But right now, the scoreboard is clear: Gemini 3 put Google back in the race and forced everyone else to rethink their roadmap. And the stock market seems to love this news.