DeepStack
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
- University of Alberta,Charles University,Czech Technical University
- Organisation type
- Academia,Academia,Academia
- Country
- Canada, Czechia
- Published
- 6 January 2017
- Authors
- Matej Moravčík, Martin Schmid, Neil Burch, Viliam Lisý, Dustin Morrill, Nolan Bard, Trevor Davis, Kevin Waugh, Michael Johanson, Michael Bowling
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Games
- Task
- Poker
- Numerical format
- FP32
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
- 2.5M
- Training data
- 25,380,000,000 tokens
Figure 3, p.9 source: https://docs.google.com/spreadsheets/d/1Kj4Q5WADcDXtUJLIOfGTCE3tGvxNczEMwyy8QtgSkHk/edit#gid=54587040&fvid=1361937389
"The turn network was trained by solving 10 million randomly generated poker turn games. These turn games used randomly generated ranges, public cards, and a random pot size (10)."
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
- 1.4 × 10¹⁹ FLOP
- How it was established
- Hardware
The largest source of compute necessary for training seems to be the data generation job on 20 GPUs. We count this towards the training compute because it requires simulation using the network. This is analogous to the AlphaGo systems simulating Go games. From p.26: "For the flop network, one million poker flop situations (from after the flop cards are dealt) were generated and solved. These situations were solved using DeepStack’s depth limited solver with the turn network used for the counter…
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Chips used
- 20
- Chip-hours
- 4,368
- Wall-clock time
- 218 hours (9.1 days)
from compute notes - around 9 days - half a year of GPU compute using 20 GPUs
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
- Record confidence
- Speculative
- Citations
- 998
first human-competitive poker AI, confirmed by website: https://www.deepstack.ai/
Sources
Where this record came from and when it was last checked.
- Reference
- DeepStack: Expert-Level Artificial Intelligence in No-Limit Poker
- Last updated
- 25 May 2026
What the numbers mean
About this model
DeepStack was published by University of Alberta,Charles University,Czech Technical University, in the country recorded as Canada, during January 2017. It comes out of an organisation categorised as academia,Academia,Academia.
It works in the domain of Games, and is recorded as performing the task of poker.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
What went into building it
The training run consumed about 1.4 × 10¹⁹ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.
It was trained on a corpus of about 25,380,000,000 tokens of text.
The reason it appears in this catalogue at all: sOTA improvement.
Answers
DeepStack — common questions
DeepStack— how many parameters does it have?
It has a parameter count of 2.5M. Figure 3, p.9 source: https://docs.google.com/spreadsheets/d/1Kj4Q5WADcDXtUJLIOfGTCE3tGvxNczEMwyy8QtgSkHk/edit#gid=54587040&fvid=1361937389. 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.
DeepStack— who created it?
It was published by University of Alberta,Charles University,Czech Technical University, based in Canada, an organisation categorised as academia,Academia,Academia.
DeepStack— when was it released?
It was published in January 2017. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
DeepStack— what is it used for?
It works in the domain of Games, and is recorded as handling the task of poker. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
DeepStack— how much compute was used to train it?
Training consumed around 1.4 × 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.
DeepStack— 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.
DeepStack— is it open source?
The licensing was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.
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.