Who makes AlphaEvolve?

AlphaEvolve is made by Google DeepMind.

Does AlphaEvolve work in the EU?

Unverified. No explicit evidence found. The agent is not publicly released; only a results and verification notebook is available. Last verified 14 Nov 2026.

How much does AlphaEvolve cost?

Pricing unknown. No public pricing found in the provided material. The agent itself is not released; only the results notebook is public. Licence: Apache-2.0 (results repository only); CC-BY-4.0 (other materials).

Where was AlphaEvolve shown?

In 3 creator videos by Matthew Berman. Most recently by Matthew Berman on 7 Nov 2026, at 7:54 in "12 months to the singularity".

AlphaEvolve

by Google DeepMind

Try it Website GitHub Paper

Gemini-powered evolutionary coding agent from Google DeepMind for algorithm and mathematical discovery.

UnverifiedResearch Agents & AutomationResearch

Updated 14 Nov 2026EU status checked 14 Nov 2026Last shown 7 Nov 2026

AlphaEvolve at a glance

What is AlphaEvolve?: AlphaEvolve is a research from Google DeepMind. Gemini-powered evolutionary coding agent from Google DeepMind for algorithm and mathematical discovery.
Who makes AlphaEvolve?: AlphaEvolve is made by Google DeepMind.
Does AlphaEvolve work in the EU?: Unverified. No explicit evidence found. The agent is not publicly released; only a results and verification notebook is available. Last verified 14 Nov 2026.
How much does AlphaEvolve cost?: Pricing unknown. No public pricing found in the provided material. The agent itself is not released; only the results notebook is public. Licence: Apache-2.0 (results repository only); CC-BY-4.0 (other materials).
Where was AlphaEvolve shown?: In 3 creator videos by Matthew Berman. Most recently by Matthew Berman on 7 Nov 2026, at 7:54 in “12 months to the singularity”.

Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.

About

AlphaEvolve is a coding agent from Google DeepMind that uses Gemini models within an evolutionary loop to propose, evaluate and refine programs. It is aimed at scientific and algorithmic discovery problems where candidate solutions can be checked automatically. It is described in the technical report "AlphaEvolve: A coding agent for scientific and algorithmic discovery" (Novikov et al., 2025, arXiv:2506.13131).

The report covers results on open mathematical problems, where AlphaEvolve found constructions that outperform the best known ones in some instances. Google has also described applying it to its own infrastructure, including hardware and data-centre efficiency. A YouTube creator cites a gain of about 1% efficiency from its use on Google's hardware architecture, which is significant at Google's scale.

The linked GitHub repository, google-deepmind/alphaevolve_results, holds only a Colab notebook with the mathematical results from Section 3 of the paper and code to verify them. It does not contain the code to run AlphaEvolve itself. Software in the repository is Apache-2.0 and other materials are CC-BY 4.0. The README states it is not an official Google product. The provided material gives no public product pricing or general-access details.

Key features

Evolutionary loop that uses Gemini models to generate and refine programs
Automated evaluation of candidate solutions for algorithmic and mathematical problems
Improved on best known constructions for some open math problems
Applied to Google infrastructure such as hardware and data-centre efficiency
Results and verification code published as a Colab notebook

3 moments

As seen on YouTube

Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.

Matthew Berman7 Nov 2026

12 months to the singularity

Mentioned as an earlier example of AI improving itself: Google pointed it at its hardware architecture and gained about 1% efficiency, which is large savings at Google scale.

7:54Open on YouTube

Matthew Berman10 Oct 2026

Everything has been solved

Google's project that looked at the code running Google's systems and found large architectural improvements. The creator says it saves Google billions of dollars per year and cites it as an example of recursive self-improvement.

15:05Open on YouTube