Google announced that it is expanding its Gemini 3 AI model to nearly 120 countries and territories, bringing the company’s most advanced artificial intelligence system to English-speaking users worldwide just two weeks after the model’s initial launch.
The expansion makes Gemini 3 available in AI Mode in Google Search for Google AI Pro and Ultra subscribers, marking a rapid global rollout for what the company describes as its most intelligent model to date. The model has already demonstrated strong early adoption, with the Gemini app reaching over 650 million monthly active users and AI Mode serving 75 million daily active users across 40 languages.
Training on Custom Chips Sets Google Apart
A key differentiator for Gemini 3 is that it was trained entirely on Google’s custom Tensor Processing Units, or TPUs, rather than relying on chips from Nvidia, according to multiple industry analysts. The seventh-generation TPU chips offer improved energy efficiency and cost advantages compared to traditional GPUs, giving Alphabet a strategic edge in the increasingly competitive AI infrastructure market.
“Reitzes notes that Gemini 3 was trained primarily on Google TPUs, giving Alphabet a price advantage,” reported Yahoo Finance, adding that outside of Nvidia’s GPUs, “the TPU is regarded as the most reliable AI chip”. The approach has attracted attention from other major tech companies, with Meta reportedly in discussions to deploy TPUs in its data centers.
Benchmark Performance Against Rivals
Early assessments suggest Gemini 3 is competitive with OpenAI’s GPT-5, which launched in August. According to Axios, “analysts, users, and industry insiders say Gemini 3’s superior benchmarks, integration into Google’s ecosystem, and cost efficiencies are pressuring OpenAI”. The model achieved a score of 1501 Elo on the LMArena Leaderboard and demonstrated PhD-level reasoning on challenging tests, including 37.5% on Humanity’s Last Exam.
Independent head-to-head testing has shown mixed results, with Gemini 3 demonstrating strengths in creative tasks and multimodal understanding, while GPT-5 maintains advantages in certain coding applications. The November 18 launch of Gemini 3 marked the first time Google shipped a new Gemini model in Search on day one, signaling the company’s commitment to integrating AI across its product ecosystem.