Agile Soccer Robot

Closed weights Google DeepMind April 2023

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Google DeepMind
Organisation type
Industry
Country
United States of America
Published
26 April 2023
Authors
Tuomas Haarnoja, Ben Moran, Guy Lever, Sandy H. Huang, Dhruva Tirumala, Markus Wulfmeier, Jan Humplik, Saran Tunyasuvunakool, Noah Y. Siegel, Roland Hafner, Michael Bloesch, Kristian Hartikainen, Arunkumar Byravan, Leonard Hasenclever, Yuval Tassa, Fereshteh Sadeghi, Nathan Batchelor, Federico Casarini, Stefano Saliceti, Charles Game, Neil Sreendra, Kushal Patel, Marlon Gwira, Andrea Huber, Nicole…

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Robotics
Task
Animal (human/non-human) imitation, Sports
Approach
Reinforcement learning

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Training data
3,140,000,000 tokens

". The get-up teacher learns to get up relatively quickly and trained in total for approximately 2.4 · 10^8 environment steps, equivalent to approximately 70 days of simulation time, or 14 hours of wall-clock time. The soccer teacher was trained for 2 · 10^9 environment steps, which took 158 hours of training, equivalent to approximately 580 days of simulated match"

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Wall-clock time
240 hours (10 days)

14+158+68 hours: "Training the get-up and soccer teachers took 14 and 158 hours (6.5 days), respectively, and distillation and self-play took 68 hours (see Appendix B for details)"

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Closed — provider access only
Model access
Unreleased
Training code
Unreleased

only video demos here https://sites.google.com/view/op3-soccer

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Why it is tracked
SOTA improvement

Likely the best bipedal soccer AI, since it's DeepMind, and related work section just discusses results involving specific soccer skills and quadruped robots: "Whether bipedal or quadrupedal, navigation represents only a fraction of animal and human capabilities. Motivated by this observation, there is a growing interest in whole body control, i.e. tasks in which the whole body is used in flexible ways to interact with the environment. Examples include climbing (Rudin et al., 2022a), getting-up…

Record confidence
Unknown
Citations
262

Sources

Where this record came from and when it was last checked.

Reference
Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning
Last updated
25 May 2026

What the numbers mean

Where it came from

Agile Soccer Robot was published by Google DeepMind, in the country recorded as United States of America, during April 2023. The publishing organisation is categorised as industry.

It works in the domain of Robotics, and is recorded as performing the task of animal (human/non-human) imitation, Sports.

Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.

What went into building it

It was trained on a corpus of about 3,140,000,000 tokens of text.

It is tracked in the underlying dataset for one reason in particular: sOTA improvement.

Answers

Agile Soccer Robot — common questions

01

Agile Soccer Robot— when was it released?

It was published in April 2023. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

02

Agile Soccer Robot— what is it used for?

It works in the domain of Robotics, and is recorded as handling the task of animal (human/non-human) imitation, Sports. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

03

Agile Soccer Robot— what GPU do I need to run it?

None. This is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

04

Agile Soccer Robot— is it open source?

No. Its weights have not been published, so it exists only as a service controlled by its owner.

05

Agile Soccer Robot— how many parameters does it have?

No parameter count has been published for it, which is why no memory or speed figure appears on this page.

06

Agile Soccer Robot— who created it?

It was published by Google DeepMind, based in United States of America, an organisation categorised as industry.

Source

Original publication

Record last updated 25 May 2026

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.