Gemini 1.0
Era 5 · Generative AI, reasoning, and agents (2022–2026) › Frontier models
- Lineage
- Neurons and networks
- Kind
- Implementation
Who
Google DeepMind
What happened
Google presents Gemini 1.0, its first model designed from the start to work with text, code, audio, images, and video at once. It comes in three sizes: Ultra, Pro, and Nano.
Why it mattered
The first model in the Gemini family. Multimodal from training onward.
The full story
On December 6, 2023, Google CEO Sundar Pichai and Google DeepMind CEO Demis Hassabis presented Gemini 1.0 as the largest and most capable model the company had built.
Google emphasized the design. According to the announcement, Gemini had been built from the ground up to be multimodal: it was pretrained from the start on different types of data, and it worked with text, code, audio, images, and video.
It came in three sizes: Ultra, the largest, for the most complex tasks; Pro, for a broad range of tasks; and Nano, designed to run on a device. That same day the Bard assistant began using a version of Gemini Pro, and the Pixel 8 Pro became the first phone built to run Gemini Nano. Ultra was at first reserved for selected customers, developers, and safety experts, and it was released on February 8, 2024.
Google claimed that Ultra, with a score of 90 percent, was the first model to outperform human experts on MMLU, a test of knowledge and problem solving across 57 subjects. The technical report spelled out the method: that figure was obtained by generating 32 chains of thought per question, whereas the 86.4 percent for GPT-4 it was compared against had been measured with five examples and a single answer.
The family advanced quickly. Gemini 1.5, in February 2024, extended the context window to a million tokens, and Gemini 2.5 Pro, in March 2025, debuted at the top of LMArena, a leaderboard based on user preferences.
Sources
Original source
- Gemini: A Family of Highly Capable Multimodal Models (arXiv:2312.11805) (opens in a new tab), tabla 2. Gemini Team, Google, 2023
Scholarship and context
- Introducing Gemini: our largest and most capable AI model (Google, December 6, 2023) (opens in a new tab). Sundar Pichai and Demis Hassabis, 2023
Further reading
How to cite
“Gemini 1.0.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/gemini-1
This entry was last reviewed on November 13, 2026.