History
Gemini 1.5 Pro reads a million tokens
Google shows a model that handles up to 1 million tokens of context, first as a limited preview for developers.

On 15 February 2024, Google announces Gemini 1.5. The first model in that line is Gemini 1.5 Pro.
What happened
- According to Google, 1.5 Pro reaches the same quality as 1.0 Ultra while it uses less compute.
- Google says the model runs consistently with up to 1 million tokens. A token is a small piece of text, often part of a word.
- The long context first comes as a limited, experimental preview for developers and enterprise customers.

Why it mattered
With that much context, you can give a model a whole case file, a long manual or a large codebase at once. Long context becomes a product feature, not only a research demo.

What you can do with it now
A lot of context is useful, but it does not prove that the model reads everything well. Test it with a question you already know the answer to, deep inside your own document.
Written afterwards on 12 November 2026, from the source below.