Google Releases Gemini 3, and Takes the Lead on Most Tests

The new model tops the standard benchmarks, comes with a new coding tool, and goes into Google Search on day one — the first time the company has put its newest AI in front of billions of people immediately.

Google released Gemini 3 today, and by the standard measures it is the best model publicly available: record scores across reasoning, math, and coding benchmarks, ahead of OpenAI's and Anthropic's current releases. It arrives with a new coding product and, more significantly, it goes into Google Search immediately rather than months later.

That last detail is the strategic one. For three years Google has been in the awkward position of inventing the technology, being beaten to market by a smaller company, and then holding its best work back from the product that billions of people actually use. Putting Gemini 3 into Search on launch day is the company deciding it is no longer playing catch-up and no longer protecting the old product from the new one.

The benchmark lead is real but should be held loosely. These scores have changed hands every few months for two years — Anthropic leads, then OpenAI, then Google — and the differences at the top are increasingly narrow and increasingly gamed. A more useful signal is what practitioners reach for after a few weeks, and that verdict takes longer.

What is not in doubt is the position Google now occupies. It has its own chips, its own data centers, the world's most-used search engine, a browser, a phone operating system, and now a model at the front of the pack. Of all the companies in this race, it is the only one that owns every layer of it.