MuZero

Closed weights DeepMind 36.9M parameters November 2019

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
DeepMind
Organisation type
Industry
Country
United Kingdom of Great Britain and Northern Ireland
Published
19 November 2019
Authors
J Schrittwieser, I Antonoglou, T Hubert, K Simonyan

What it does

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

Domain
Games
Task
Atari
Approach
Self-supervised 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.

Parameters
36.9M

Both the representation and dynamics function use the same architecture asAlphaZero, but with 16 instead of20 residual blocks [15]. We use 3x3 kernels and 256 hidden planes for each convolution. Previous downsampling: • 1 convolution with stride 2 and 128 output planes, output resolution 48x48.• 2 residual blocks with 128 planes• 1 convolution with stride 2 and 256 output planes, output resolution 24x24.• 3 residual blocks with 256 planes.• Average pooling with stride 2, output resolution …

Training data
12,288,000,000 tokens

Table 1 https://arxiv.org/pdf/1911.08265.pdf

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

Training compute
4.8 × 10¹⁹ FLOP

third-generation Google Cloud TPU (For each board game, we used 16 TPUs for training and 1000 TPUs for self-play) For each game in Atari, we used 8 TPUs for training and 32 TPUs for self-play Training for 12 hours (for Atari) Data from Parameter, Compute and Data Trends in Machine Learning Google v3 TPU: 1.23E+14 FLOP/s (although with the caveat that it might be not applicable) Utilization rate In LaMDA: Language Models for Dialog Applications, they report for TPU V3: 56.5% Calculations for Ata…

How it was established
Hardware

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

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
Highly cited,SOTA improvement

"When evaluated on 57 different Atari games - the canonical video game environment for testing AI techniques, in which model-based planning approaches have historically struggled - our new algorithm achieved a new state of the art. "

Record confidence
Speculative
Citations
2,491

Sources

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

Reference
Mastering Atari Go Chess and Shogi by Planning with a Learned Model
Last updated
25 May 2026

What the numbers mean

What this model is

MuZero was published by DeepMind, in the country recorded as United Kingdom of Great Britain and Northern Ireland, during November 2019. The publishing organisation is categorised as industry.

It works in the domain of Games, and is recorded as performing the task of atari.

Its weights were never published, so it can only be reached through its provider. No graphics card changes that.

What went into building it

Producing it required arithmetic totalling around 4.8 × 10¹⁹ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.

The training set ran to roughly 12,288,000,000 tokens of text.

Its inclusion criterion: highly cited,SOTA improvement.

Answers

MuZero — common questions

01

MuZero— how much compute was used to train it?

Training consumed around 4.8 × 10¹⁹ FLOP. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.

02

MuZero— 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.

03

MuZero— is it open source?

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

04

MuZero— how many parameters does it have?

It has a parameter count of 36.9M. Both the representation and dynamics function use the same architecture asAlphaZero, but with 16 instead of20 residual blocks [15]. We use 3x3 kernels and 256 hidden planes for each convolution. Previous downsampling: • 1 convolution with stride 2 and 128 output planes, output resolution 48x48.• 2 residual blocks with 128 planes• 1 convolution with stride 2 and 256 output planes, output resolution 24x24.• 3 residual blocks with 256 planes.• Average pooling with stride 2, output resolution 12x12.• 3 residual blocks with 256 planes.• Average pooling with stride 2, output resolution 6x6. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

05

MuZero— who created it?

It was published by DeepMind, based in United Kingdom of Great Britain and Northern Ireland, an organisation categorised as industry.

06

MuZero— when was it released?

It was published in November 2019. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

07

MuZero— what is it used for?

It works in the domain of Games, and is recorded as handling the task of atari. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

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.