RETRO-7B
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
- 7 February 2022
- Authors
- Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George van den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, Diego de Las Casas, Aurelia Guy, Jacob Menick, Roman Ring, Tom Hennigan, Saffron Huang, Loren Maggiore, Chris Jones, Albin Cassirer, Andy Brock, Michela Paganini, Geoffrey Irving, Oriol Vinyals, Simon Osindero,Karen Simonyan, …
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling/generation, Language modeling
- 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
- 7.5B
- Training data
- 419,430,400,000 tokens
- Batch size
- 2,097,152
"Retro provides a constant gain for models ranging from 150M to 7B parameters, and Retro can be improved at evaluation time by increasing the database size and the number of retrieved neighbours. "
"we train for 419,430,400,000 training tokens" ~= 315B words.
1024 * 2048
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
- 1.7 × 10²² FLOP
- How it was established
- Operation counting
C=6ND = 6 * 7e9 * 400e9 = 1.7e22
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
- Record confidence
- Confident
- Citations
- 1,665
- Benchmark data
- RETRO-7B
"Our largest model obtains state-of-the-art results on a range of downstream evaluation datasets including Wikitext103"
Sources
Where this record came from and when it was last checked.
- Reference
- Improving language models by retrieving from trillions of tokens
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
RETRO-7B was published by DeepMind, in United Kingdom of Great Britain and Northern Ireland, in February 2022. The organisation is categorised as industry.
It works in Language, and is recorded as doing language modeling/generation, Language modeling.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
What went into building it
Training it took roughly 1.7 × 10²² FLOP of computation — a measure of what producing the model cost, not of how fast it answers.
Around 419,430,400,000 tokens went into training it.
The reason it appears in this catalogue at all is sOTA improvement.
Answers
RETRO-7B — common questions
What GPU do I need to run RETRO-7B?
None. RETRO-7B 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.
Is RETRO-7B open source?
No. RETRO-7B has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does RETRO-7B have?
RETRO-7B has 7.5B parameters. "Retro provides a constant gain for models ranging from 150M to 7B parameters, and Retro can be improved at evaluation time by increasing the database size and the number of retrieved neighbours. ". 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.
Who created RETRO-7B?
RETRO-7B was published by DeepMind, based in United Kingdom of Great Britain and Northern Ireland, categorised as industry.
When was RETRO-7B released?
RETRO-7B was published in February 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
What is RETRO-7B used for?
RETRO-7B works in Language, and is recorded as handling language modeling/generation, Language modeling. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
How much compute was used to train RETRO-7B?
Around 1.7 × 10²² 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.
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.