GOAT

Closed weights DeepMind 3.5M parameters July 2021

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
27 July 2021
Authors
Open-Ended Learning Team*, Adam Stooke, Anuj Mahajan, Catarina Barros, Charlie Deck, Jakob Bauer, Jakub Sygnowski, Maja Trebacz, Max Jaderberg, Michael Mathieu, Nat McAleese, Nathalie Bradley-Schmieg, Nathaniel Wong, Nicolas Porcel, Roberta Raileanu, Steph Hughes-Fitt, Valentin Dalibard and Wojciech Marian Czarnecki

What it does

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

Domain
Games
Task
Open ended play
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
3.5M

estimate described here: https://docs.google.com/document/d/1S9xZyCeITDOs-P1W_-liNW0WgVN-OLsSudVrPXMaLqw/edit?usp=sharing

Training data
798,720,000,000,000 tokens

Figure 16 shows steps per generation and agent. In total there are 1.5e10 + 4.0e10 + 2.5e10 + 1.1e11 + 2e11 = 3.9e11 steps per agent.

Batch size
64

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
2.4 × 10²² FLOP

[Final calculation] (8 TPUs) * (1.23e14 FLOP/TPU-s) * (0.1 utilization) / (50k steps/s) = 1.968e9 FLOP/step (32 agents) * (383B steps/agent) * (1.968e9 FLOP/step) = 2.412e22 FLOPs ========================== NOTES BELOW 6.1: Each agent is trained using 8 TPUv3s and consumes approximately 50,000 agent steps (observations) per second. Multiple agents interacting probably mean a fairly low utilization rate, so let’s assume 0.10 8 * 1.23e14 * 0.1 / 50k = 1.968e9 FLOPs per step The paper doesn’t s…

How it was established
Hardware

The training run

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

Training hardware
Google TPU v3
Compute cost
$84,800

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.

Foundation model
Yes
Why it is tracked
SOTA improvement

likely qualitatively SOTA I do not see any standard benchmarks that they are claiming SOTA on

Record confidence
Speculative
Citations
228

Sources

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

Reference
Open-Ended Learning Leads to Generally Capable Agents
Last updated
25 May 2026

What the numbers mean

About this model

GOAT was published by DeepMind, in the country recorded as United Kingdom of Great Britain and Northern Ireland, during July 2021. The publishing organisation is categorised as industry.

It works in the domain of Games, and is recorded as performing the task of open ended play.

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

What went into building it

The training run consumed about 2.4 × 10²² FLOP, on hardware recorded as Google TPU v3. That figure measures what producing the model cost, and has no bearing on how fast it answers.

The training set ran to roughly 798,720,000,000,000 tokens of text.

Its inclusion criterion: sOTA improvement.

Answers

GOAT — common questions

01

GOAT— is it open source?

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

02

GOAT— how many parameters does it have?

It has a parameter count of 3.5M. estimate described here: https://docs.google.com/document/d/1S9xZyCeITDOs-P1W_-liNW0WgVN-OLsSudVrPXMaLqw/edit?usp=sharing. 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.

03

GOAT— who created it?

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

04

GOAT— when was it released?

It was published in July 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

05

GOAT— what is it used for?

It works in the domain of Games, and is recorded as handling the task of open ended play. 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.

06

GOAT— how much compute was used to train it?

Training consumed around 2.4 × 10²² FLOP, on hardware recorded as Google TPU v3. 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.

07

GOAT— 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.

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.