According to Business Insider, OpenAI CEO Sam Altman sent an internal memo declaring a “code red” situation following the early November launch of Google’s Gemini 3, calling for rapid improvements to ChatGPT. The new Gemini model scored 37.5% on the Humanity’s Last Exam benchmark without tool use, beating GPT-5.1’s score of 26.5%. On the LMArena leaderboard, Gemini 3 leads in overall user ratings, with ChatGPT-5.1 ranking about 300 points behind in third. Pricing is competitive: ChatGPT Plus is $20/month, while Google AI Pro is $19.99/month and free for students for a year, with top-tier plans at $200 and $249.99 monthly respectively. Principal scientist Mayank Kejriwal called Gemini 3 the biggest leap in LLMs this year, praising its unified processing of text, video, audio, and code.
The benchmark battle
Here’s the thing about benchmarks: they’re not everything, but they’re the scoreboard everyone looks at. And right now, Google is putting points on the board. A double-digit lead on a major test like Humanity’s Last Exam is a big deal. It’s the kind of gap that makes product managers sweat and gets CEOs writing “code red” memos. But it’s the LMArena leaderboard that’s maybe more telling. That’s a blind test where real users rate responses without knowing which model they’re talking to. Gemini 3 leading there suggests the improvements aren’t just academic; they’re noticeable to people just asking questions. That’s how you change perceptions.
Integration is everything
Now, raw power is one thing. But where you live is another. And this is Google’s classic home-field advantage. Gemini 3 isn’t just an app you open; it’s baked into Search, Gmail, Docs, and the whole Workspace suite. For the billions of people already inside Google’s world, that’s a frictionless win. OpenAI‘s strength has been its wide-open platform approach with plugins and APIs, which is fantastic for developers. But Google is targeting the everyday productivity user. Click “AI mode” in your search bar? That’s a genius move for mass adoption. It makes AI feel less like a separate tool and more like a natural extension of what you’re already doing.
The pricing and psychology war
Look at the pricing. They’re basically mirror images of each other, right down to the penny on the mid-tier plans. This is a direct, head-on fight for the same subscribers. But Google throwing in a free year of AI Pro for all students? That’s a long-game play to hook the next generation of users. It’s aggressive. And the “code red” leak from OpenAI is perhaps the most significant data point of all. For years, OpenAI set the pace and Google seemed to be playing catch-up. That dynamic has officially flipped. The pressure is now on OpenAI to answer with a major update of its own. The race just got a lot more interesting.
What unified reasoning really means
Kejriwal’s point about unified processing is the philosophical core of this. The idea of one model seamlessly handling text, audio, video, and code—like a human brain—is the AGI holy grail. Gemini 3’s “Thinking” mode in search tries to surface that reasoning. That’s the vision: an AI that doesn’t just retrieve information but truly understands context across different mediums. If Google is pulling ahead here, it’s not just about better answers to trivia questions. It’s about building a more generally intelligent, cohesive assistant. The question is, can OpenAI’s modular, partnership-driven approach keep up with that kind of deeply integrated, unified vision? The next six months will give us the answer.
