AlphaFold-Multimer
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,DeepMind
- Organisation type
- Industry,Industry
- Country
- United States of America, United Kingdom of Great Britain and Northern Ireland
- Published
- 4 October 2021
- Authors
- Richard Evans, Michael O’Neill, Alexander Pritzel, Natasha Antropova, Andrew Senior, Tim Green, Augustin Žídek, Russ Bates, Sam Blackwell, Jason Yim, Olaf Ronneberger, Sebastian Bodenstein, Michal Zielinski, Alex Bridgland, Anna Potapenko, Andrew Cowie, Kathryn Tunyasuvunakool, Rishub Jain, Ellen Clancy, Pushmeet Kohli, John Jumper and 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
- Base model
- AlphaFold 2
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
- 56,573,952 tokens
- Epochs
- 68
See: https://www.rcsb.org/stats/growth/growth-released-structures for 2018 "We train the model to convergence (approximately 10M samples, for 2 weeks) across 128 TPUv3 cores with a batch size of 1 per TPU core. Then we halve the learning rate and double the number of sequences fed into the MSA stack before running two separate fine-tuning stages (one further day of training each)" 10000000/147328 ~ 68 epochs
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
- 4.4 × 10²¹ FLOP
- How it was established
- Hardware
Section: 2.5. Training Regimen "We train the model to convergence (approximately 10M samples, for 2 weeks) across 128 TPUv3 cores [...]. Then we [...] run two separate fine-tuning stages (one further day of training each)" Assuming: FP16 and utilization 0.4 Calculation: (14+2) days * 24 hours/day * 60 min/hour * 60 sec/min * (128 TPU cores/2 cores per chip) * 1.23e14 FLOP/s per chip * 0.4 utilization = 4.35e21 FLOPs
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
- 64
- Chip-hours
- 24,576
- Wall-clock time
- 384 hours (16 days)
- Power draw
- 58.1 kW
- Compute cost
- $7,966
Section: 2.5. Training Regimen "We train the model to convergence (approximately 10M samples, for 2 weeks) across 128 TPUv3 cores [...]. Then we [...] run two separate fine-tuning stages (one further day of training each)"
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 (unrestricted)
- Training code
- Unreleased
While the AlphaFold code is licensed under the Apache 2.0 License, the AlphaFold parameters and CASP15 prediction data are made available under the terms of the CC BY 4.0 license https://github.com/google-deepmind/alphafold
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
- Highly cited,SOTA improvement
- Record confidence
- Confident
- Citations
- 2,694
"On a benchmark dataset of 17 heterodimer proteins without templates (introduced in [2]) we achieve at least medium accuracy (DockQ [3] ≥ 0.49) on 14 targets and high accuracy (DockQ ≥ 0.8) on 6 targets, compared to 9 targets of at least medium accuracy and 4 of high accuracy for the previous state of the art system (an AlphaFold-based system from [2])" "For heteromeric interfaces we successfully predict the interface (DockQ ≥ 0.23) in 67% of cases, and produce high accuracy predictions (DockQ …
Sources
Where this record came from and when it was last checked.
- Reference
- Protein complex prediction with AlphaFold-Multimer
- Last updated
- 1 January 2026
What the numbers mean
Background
AlphaFold-Multimer was published by Google DeepMind,DeepMind, in the country recorded as United States of America, during October 2021. The publishing organisation is categorised as industry,Industry.
It works in the domain of Biology, and is recorded as performing the task of protein folding prediction, Proteins.
Rather than being trained from scratch, it is derived from AlphaFold 2. That is the usual way a specialised model is produced.
Published weights mean the model runs on your machine rather than someone else's, which is what makes the hardware question below answerable at all.
How it was trained
Training it took a computation budget of roughly 4.4 × 10²¹ 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 56,573,952 tokens of text.
It is tracked in the underlying dataset for one reason in particular: highly cited,SOTA improvement.
Answers
AlphaFold-Multimer — common questions
AlphaFold-Multimer— what is it used for?
It works in the domain of Biology, and is recorded as handling the task of protein folding prediction, Proteins. These are the areas it was designed around; they describe intent rather than a hard boundary.
AlphaFold-Multimer— where can I download it?
The weights are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.
AlphaFold-Multimer— how much compute was used to train it?
Training consumed around 4.4 × 10²¹ 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.
AlphaFold-Multimer— what GPU do I need to run it?
We cannot say. It 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.
AlphaFold-Multimer— is it open source?
Its weights are published, so it 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.
AlphaFold-Multimer— how many parameters does it have?
No parameter count has been published for it, which is why no memory or speed figure appears on this page.
AlphaFold-Multimer— who created it?
It was published by Google DeepMind,DeepMind, based in United States of America, an organisation categorised as industry,Industry.
AlphaFold-Multimer— when was it released?
It was published in October 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.
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.