Multilingual DNN
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
- Organisation type
- Industry
- Country
- United States of America
- Published
- 26 May 2013
- Authors
- G. Heigold, Vincent Vanhoucke, A. Senior, Patrick Nguyen, Marc'Aurelio Ranzato, M. Devin, J. Dean
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Speech
- Task
- Speech recognition (ASR)
- Approach
- Supervised
- 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
- 206.9M
- Training data
- 3,103,200,000 tokens
"The input for the DNN is eleven contiguous frames of 40-dimensional log-filterbank features. The DNN consists of four hidden layers each with 2560 nodes" Network structure: 3 multilingual shared layers, 1 language specific hidden layer + output layer (Figure 2) Language specific layer output sizes: 1600, 3300, 2900, 5700, 3500, 5500, 6200, 4700, 5100, 4900, 3700 (Table 1) Shared: 11*40*2560+2560*2560+2560*2560=14233600 Language heads: 11*2560*2560+2560*1600+2560*3300+2560*2900+2560*5700+2560*35…
Trained on 80+100+220+270+920+1140+1450+1460+1490+1490=8620h of speech data (Table 1) Conversion to words using an estimate of 150 wpm: 8620*60*150=77580000 words
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
- 672 hours (28 days)
"increased training time of roughly four weeks" 4*7*24=672 hours of training
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,Training cost
- Record confidence
- Confident
I don't see any standard benchmarks where they would claim SOTA results
Sources
Where this record came from and when it was last checked.
- Reference
- Multilingual acoustic models using distributed deep neural networks
- Last updated
- 28 November 2025
What the numbers mean
Background
Multilingual DNN was published by Google, in the country recorded as United States of America, during May 2013. It comes out of an organisation categorised as industry.
It works in the domain of Speech, and is recorded as performing the task of speech recognition (ASR).
This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.
What went into building it
Training consumed a corpus of around 3,103,200,000 tokens of text.
The reason it appears in this catalogue at all: sOTA improvement,Training cost.
Answers
Multilingual DNN — common questions
Multilingual DNN— what is it used for?
It works in the domain of Speech, and is recorded as handling the task of speech recognition (ASR). A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.
Multilingual DNN— 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.
Multilingual DNN— is it open source?
The licensing was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.
Multilingual DNN— how many parameters does it have?
It has a parameter count of 206.9M. "The input for the DNN is eleven contiguous frames of 40-dimensional log-filterbank features. The DNN consists of four hidden layers each with 2560 nodes" Network structure: 3 multilingual shared layers, 1 language specific hidden layer + output layer (Figure 2) Language specific layer output sizes: 1600, 3300, 2900, 5700, 3500, 5500, 6200, 4700, 5100, 4900, 3700 (Table 1) Shared: 11*40*2560+2560*2560+2560*2560=14233600 Language heads: 11*2560*2560+2560*1600+2560*3300+2560*2900+2560*5700+2560*3500+2560*5500+2560*6200+2560*4700+2560*5100+2560*4900+2560*3700=192665600 Total: 14233600+192665600=206899200=2e8. 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.
Multilingual DNN— who created it?
It was published by Google, based in United States of America, an organisation categorised as industry.
Multilingual DNN— when was it released?
It was published in May 2013. 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.