AlphaStar
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
- 30 October 2019
- Authors
- Oriol Vinyals,Igor Babuschkin,Wojciech M. Czarnecki,Michaël Mathieu,Andrew Dudzik,Junyoung Chung,David H. Choi,Richard Powell,Timo Ewalds,Petko Georgiev,Junhyuk Oh,Dan Horgan,Manuel Kroiss,Ivo Danihelka,Aja Huang,Laurent Sifre,Trevor Cai,John P. Agapiou,Max Jaderberg,Alexander S. Vezhnevets,Rémi Leblond,Tobias Pohlen,Valentin Dalibard,David Budden,Yury Sulsky,James Molloy,Tom L. Paine,Caglar Gulce…
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Games
- Task
- StarCraft
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
- 139M
- Training data
- tokens
AlphaStar has 139 million weights, but only 55 million weights are required during inference.
Multiple data types. First supervised learning, then other stuff
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
- 1.1 × 10²³ FLOP
- How it was established
- Hardware
(Estimate by James Sanders, checked by Robi Rahman) Fig 6 indicates that the learner uses 16 TPU "devices" which are 128 TPU cores total, which matches 4 TPUs per device, and 2 cores per TPU. Fig 6 indicates that 64 TPUs are used for training, and 64 are used for inference. (128 TPUs)*(12 agents)*(44 days)*(123 TFLOPS)*(0.3 utilization) = 2.15e23 FLOP (total), of which 1.07e23 FLOP are for training.
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
- Chips used
- 384
- Chip-hours
- 405,504
- Wall-clock time
- 1,056 hours (44 days)
- Power draw
- 354.2 kW
- Compute cost
- $130,654
"Each agent was trained using 32 third-generation tensor processing units (TPUs) over 44 days"
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
- Open source
Apache 2.0, training tools: https://github.com/google-deepmind/alphastar training instructions here: https://github.com/google-deepmind/alphastar/blob/main/alphastar/unplugged/README.md
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Frontier model
- Yes
- Why it is tracked
- Discretionary
- Record confidence
- Confident
- Citations
- 4,093
Sources
Where this record came from and when it was last checked.
- Reference
- Grandmaster level in StarCraft II using multi-agent reinforcement learning
- Last updated
- 1 January 2026
What the numbers mean
Background
AlphaStar was published by DeepMind, in the country recorded as United Kingdom of Great Britain and Northern Ireland, during October 2019. It comes out of an organisation categorised as industry.
It works in the domain of Games, and is recorded as performing the task of starCraft.
This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.
How it was trained
The training run consumed about 1.1 × 10²³ 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.
Its inclusion criterion: discretionary.
Answers
AlphaStar — common questions
AlphaStar— what is it used for?
It works in the domain of Games, and is recorded as handling the task of starCraft. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
AlphaStar— how much compute was used to train it?
Training consumed around 1.1 × 10²³ 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.
AlphaStar— 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.
AlphaStar— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
AlphaStar— how many parameters does it have?
It has a parameter count of 139M. AlphaStar has 139 million weights, but only 55 million weights are required during inference. 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.
AlphaStar— who created it?
It was published by DeepMind, based in United Kingdom of Great Britain and Northern Ireland, an organisation categorised as industry.
AlphaStar— when was it released?
It was published in October 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.
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.