AlphaFold

Closed weights DeepMind 16.3M parameters January 2020

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
15 January 2020
Authors
Andrew W. Senior, Richard Evans, John Jumper, James Kirkpatrick, Laurent Sifre, Tim Green, Chongli Qin, Augustin Žídek, Alexander W. R. Nelson, Alex Bridgland, Hugo Penedones, Stig Petersen, Karen Simonyan, Steve Crossan, Pushmeet Kohli, David T. Jones, David Silver, Koray Kavukcuoglu, Demis Hassabis

What it does

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

Domain
Biology
Task
Protein folding prediction, Proteins
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
16.3M

"Neural network hyperparameters" section of https://www.nature.com/articles/s41586-019-1923-7: “7 × 4 Blocks with 256 channels, cycling through dilations 1, 2, 4, 8” “48 × 4 Blocks with 128 channels, cycling through dilations 1, 2, 4, 8” "Distogram prediction" section: "For the final layer, a position-specific bias was used" Extended Data Fig.1 (b): Shows that each block consists of 9 layers: (1) Batch norm (2) Elu (3) Project down (halves number of dimensions) (4) Batch norm (5) Elu (6) 3x3 …

Training data
6,622,252,080 tokens

Training Domains: 29,427 Average Residues per Domain: 100 Data Points per Domain: 100 × 100 = 10,000 Total Data Points = 29,427 × 10,000 = 294,270,000 ≈ 3.0 × 10⁸

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 × 10²⁰ FLOP

Estimated in the blogpost below https://www.lesswrong.com/posts/wfpdejMWog4vEDLDg/ai-and-compute-trend-isn-t-predictive-of-what-is-happening "AlphaFold: they say they trained on GPU and not TPU. Assuming V100 GPU, it's 5 days * 24 hours/day * 3600 sec/hour * 8 V100 GPU * 100*10^12 FLOP/s * 33% actual GPU utilization = 10^20 FLOP."

How it was established
Hardware,Third-party estimation

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
120 hours

"Training time: about 5 days for 600,000 steps"

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
SOTA improvement,Highly cited

"AlphaFold represents a considerable advance in protein-structure prediction." [Abstract]

Record confidence
Speculative
Citations
2,773

Sources

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

Reference
Improved protein structure prediction using potentials from deep learning
Last updated
28 November 2025

What the numbers mean

What this model is

AlphaFold was published by DeepMind, in the country recorded as United Kingdom of Great Britain and Northern Ireland, during January 2020. The publishing organisation is categorised as industry.

It works in the domain of Biology, and is recorded as performing the task of protein folding prediction, Proteins.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

Training and provenance

Training it took a computation budget of roughly 1 × 10²⁰ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.

The training set ran to roughly 6,622,252,080 tokens of text.

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

Answers

AlphaFold — common questions

01

AlphaFold— how much compute was used to train it?

Training consumed around 1 × 10²⁰ 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

AlphaFold— 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

AlphaFold— is it open source?

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

04

AlphaFold— how many parameters does it have?

It has a parameter count of 16.3M. "Neural network hyperparameters" section of https://www.nature.com/articles/s41586-019-1923-7: “7 × 4 Blocks with 256 channels, cycling through dilations 1, 2, 4, 8” “48 × 4 Blocks with 128 channels, cycling through dilations 1, 2, 4, 8” "Distogram prediction" section: "For the final layer, a position-specific bias was used" Extended Data Fig.1 (b): Shows that each block consists of 9 layers: (1) Batch norm (2) Elu (3) Project down (halves number of dimensions) (4) Batch norm (5) Elu (6) 3x3 kernel with dilation (7) Batch norm (8) Elu (9) Project up (doubles number of dimensions) Dilations don't change the number of parameters in each filter Assuming that projection layers are convolutional layers with 1x1 kernels Parameter estimate for each layer in a 256 channel block: (1) 256*2 = 512 (2) 0 (3) 1*1*256*128 = 32768 (4) 128*2 = 256 (5) 0 (6) 3*3*128*128 = 147456 (7) 128*2 = 256 (8) 0 (9) 1*1*128*256 + 256 = 33024 Total = 214272 Parameter estimate for each layer in a 128 channel block: (1) 128*2 = 256 (2) 0 (3) 1*1*128*64 = 8192 (4) 64*2 = 128 (5) 0 (6) 3*3*64*64 = 36864 (7) 64*2 = 128 (8) 0 (9) 1*1*64*128 + 128 = 8320 Total = 53897 Estimate total network = 7*4*214272 + 48*4*53897 = 5992616 + 10348224 = 16340840 ~ 16e6 Within a factor of 2 of the estimate of 21M parameters stated in: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7305407/ [Previous approximation: 7 * 4 * 256 * 3 * 3 * 256 + 48 * 4 * 128 * 3 * 3 * 128 = 44826624]. 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

AlphaFold— who created it?

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

06

AlphaFold— when was it released?

It was published in January 2020. 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

AlphaFold— what is it used for?

It works in the domain of Biology, and is recorded as handling the task of protein folding prediction, Proteins. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Source

Original publication

Record last updated 28 November 2025

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.