AlphaFold 3
No estimate
No hardware requirements for this model
This model's weights are open, but no parameter count has been published for it. Every memory and speed figure starts from that number, so we would rather show nothing than a fabricated estimate.
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
- Google DeepMind,Isomorphic Labs
- Organisation type
- Industry,Industry
- Country
- United States of America, United Kingdom of Great Britain and Northern Ireland
- Published
- 8 May 2024
- Authors
- Josh Abramson, Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, Akvilė Žemgulytė, Eirini Arvaniti, Charles Beattie, Ottavia Bertolli, Alex Bridgland, Alexey Cherepanov, Miles Congre…
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Biology
- Task
- Protein folding prediction, Antibody property prediction, Protein-ligand contact prediction, RNA structure prediction, Protein interaction prediction
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.
- Training data
- 29,491,200,000 tokens
- Epochs
- 1
from https://www.biorxiv.org/content/10.1101/2024.11.19.624167v2.full.pdf "As a comparison AlphaFol3 trained a similar architecture for nearly 150k steps with a batch size of 256" from supplementary materials "The model is trained with a batch size of 256" 150000 steps * 256 sequences per batch * 384 tokens per batch [at initial training stage, Table 6, supplementary materials] = 14 745 600 000 tokens from supplementary materials Table 6 fine-tuning took exactly the same amount of GPU-hours -…
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
- 4.1 × 10²² FLOP
- How it was established
- Hardware
256 GPUs * 480 hours [see training time notes] * 3600 sec / hour * 312000000000000 FLOP / GPU / sec * 0.3 [assumed utilization] = 4.1405645e+22 FLOP
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
- NVIDIA A100
- Chips used
- 256
- Wall-clock time
- 480 hours (20 days)
- Power draw
- 202.3 kW
supplementary materials Table 6 20 days *24 hours = 480 hours
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
- Open — downloadable
- Model access
- Open weights (non-commercial)
- Training code
- Unreleased
Inference code CC-BY-NC-SA 4.0 weights - permissions given after approval (only non-commercial use) "AlphaFold 3 will be available as a non-commercial usage only server at https://www.alphafoldserver.com, with restrictions on allowed ligands and covalent modifications. Pseudocode describing the algorithms is available in the Supplementary Information. Code is not provided"
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
- Confident
Figure 4 + "we achieve a higher average performance than RoseTTAFold2NA and AIchemy_RNA27 (the best AI-based submission in CASP1518,31)" "Even so, AF3 greatly outperforms classical docking tools such as Vina37,38 even while not using any structural inputs (Fisher’s exact test, P = 2.27 × 10−13) and greatly outperforms all other true blind docking like RoseTTAFold All-Atom " also is on top of the CAMEO leaderboard https://cameo3d.org/
Sources
Where this record came from and when it was last checked.
- Reference
- Accurate structure prediction of biomolecular interactions with AlphaFold 3
- Last updated
- 28 November 2025
What the numbers mean
Where it came from
AlphaFold 3 was published by Google DeepMind,Isomorphic Labs, in United States of America, in May 2024. industry,Industry is the category the publisher falls under.
It works in Biology, and is recorded as doing protein folding prediction, Antibody property prediction, Protein-ligand contact prediction, RNA structure prediction, Protein interaction prediction.
The weights being open is what puts this page in the calculator rather than only in the catalogue: it is a model you can actually hold.
How it was trained
The training run consumed about 4.1 × 10²² FLOP, on NVIDIA A100. That figure describes the cost of creating it and has no bearing on how quickly it generates text.
The training set ran to roughly 29,491,200,000 tokens.
The reason it appears in this catalogue at all is sOTA improvement.
Answers
AlphaFold 3 — common questions
What is AlphaFold 3 used for?
AlphaFold 3 works in Biology, and is recorded as handling protein folding prediction, Antibody property prediction, Protein-ligand contact prediction, RNA structure prediction, Protein interaction prediction. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
Where can I download AlphaFold 3?
The weights for AlphaFold 3 are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.
How much compute was used to train AlphaFold 3?
Around 4.1 × 10²² FLOP, on NVIDIA A100. 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.
What GPU do I need to run AlphaFold 3?
We cannot say. AlphaFold 3 has open weights, but no parameter count has been published for it, and every memory and speed calculation starts from that number. We would rather show nothing than a fabricated estimate.
Is AlphaFold 3 open source?
Its weights are published, so AlphaFold 3 can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.
How many parameters does AlphaFold 3 have?
No parameter count has been published for AlphaFold 3, which is why no memory or speed figure appears on this page.
Who created AlphaFold 3?
AlphaFold 3 was published by Google DeepMind,Isomorphic Labs, based in United States of America, categorised as industry,Industry.
When was AlphaFold 3 released?
AlphaFold 3 was published in May 2024. 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.