Transformer
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
- Google Research,Google Brain
- Organisation type
- Industry,Industry
- Country
- United States of America
- Published
- 12 June 2017
- Authors
- Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, Illia Polosukhin
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Translation
- Approach
- Self-supervised learning
- 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
- 213M
- Training data
- 832,000,000 tokens
- Epochs
- 3
This page suggests the transformer has 213M parameters. "Although there are others architectures that make use of attention layers, none achieves so good results so fast. Not only that, but the only model that can compite against Transformer is the Slicenet22, proposed just fifteen days before. It takes much longer to train, due to the huge amount of parameters it requires (348 million against the 213 millions of Transformer), and the BLEU scores it achieves are slightly worse on average. In sh…
"We trained on the standard WMT 2014 English-German dataset consisting of about 4.5 million sentence pairs. Sentences were encoded using byte-pair encoding [3], which has a shared source-target vocabulary of about 37000 tokens. For English-French, we used the significantly larger WMT 2014 English-French dataset consisting of 36M sentences and split tokens into a 32000 word-piece vocabulary " In total, this is 40.5 million sentence-pairs. Assuming each sentence pair is 15-20 words in each langua…
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
- 7.4 × 10¹⁸ FLOP
- How it was established
- Hardware,Third-party estimation
"The model was trained during 300000 steps, roughly 3.5 days, using 8 NVIDIA P100 GPUs." source: https://ricardokleinklein.github.io/2017/11/16/Attention-is-all-you-need.html NVIDIA Tesla P100 has 9.3 teraFLOPS single-precision performance source: https://www.nvidia.com/en-gb/data-center/tesla-p100/ We assume 0.33 utilization performance, in line with OpenAI's "AI and compute" article source: https://openai.com/blog/ai-and-compute/ 9.3*10^12 FLOP / GPU / sec * 8 GPUs * 3.5 days * 24 hour /…
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 P100
- Chips used
- 8
- Chip-hours
- 672
- Wall-clock time
- 84 hours
- Power draw
- 4.2 kW
- Compute cost
- $438
We trained our models on one machine with 8 NVIDIA P100 GPUs. For our base models using the hyperparameters described throughout the paper, each training step took about 0.4 seconds. We trained the base models for a total of 100,000 steps or 12 hours. For our big models,(described on the bottom line of table 3), step time was 1.0 seconds. The big models were trained for 300,000 steps (3.5 days).
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
- Highly cited,Historical significance
- Record confidence
- Confident
- Citations
- 159,251
The original transformer
Sources
Where this record came from and when it was last checked.
- Reference
- Attention Is All You Need
- Last updated
- 1 January 2026
What the numbers mean
Where it came from
Transformer was published by Google Research,Google Brain, in United States of America, in June 2017. The organisation is categorised as industry,Industry.
It works in Language, and is recorded as doing translation.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
Training and provenance
Producing it required around 7.4 × 10¹⁸ FLOP of arithmetic, on NVIDIA P100, which is a statement about the training budget rather than about inference.
Around 832,000,000 tokens went into training it.
It is tracked in the underlying dataset for one reason in particular: highly cited,Historical significance.
Answers
Transformer — common questions
How much compute was used to train Transformer?
Around 7.4 × 10¹⁸ FLOP, on NVIDIA P100. 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 Transformer?
None. Transformer 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 Transformer open source?
No. Transformer has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does Transformer have?
Transformer has 213M parameters. This page suggests the transformer has 213M parameters. "Although there are others architectures that make use of attention layers, none achieves so good results so fast. Not only that, but the only model that can compite against Transformer is the Slicenet22, proposed just fifteen days before. It takes much longer to train, due to the huge amount of parameters it requires (348 million against the 213 millions of Transformer), and the BLEU scores it achieves are slightly worse on average. In short, up to date it offers no profit over Transformer." https://ricardokleinklein.github.io/2017/11/16/Attention-is-all-you-need.html. 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 Transformer?
Transformer was published by Google Research,Google Brain, based in United States of America, categorised as industry,Industry.
When was Transformer released?
Transformer was published in June 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.
What is Transformer used for?
Transformer works in Language, and is recorded as handling translation. These are the areas it was designed around; they describe intent rather than a hard boundary.
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.