ReLU-Speech

Closed weights Google,University of Toronto,New York University (NYU) 101.7M parameters May 2013

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,University of Toronto,New York University (NYU)
Organisation type
Industry,Academia,Academia
Country
United States of America, Canada
Published
26 May 2013
Authors
Matthew D. Zeiler, Marc'Aurelio Ranzato, R. Monga, Mark Z. Mao, K. Yang, Quoc V. Le, Patrick Nguyen, A. Senior, Vincent Vanhoucke, J. Dean, Geoffrey E. Hinton

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Speech
Task
Speech recognition (ASR)

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
101.7M

"The overall input dimensionality is 1040," "All layers of our networks have 2560 hidden units " "used to generate 7969 context-dependent tied acoustic states" Largest model: 12 hidden layers (Fig 4) Parameters: 1040*2560+12*2560*2560+2560*7969=101706240

Training data
tokens

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
1.3 × 10¹⁷ FLOP

"across 4 machines using up to 4 CPUs each" CPU model not specified, I assumed a Sandy Bridge with 16 FLOP/cycle and 3.3GhZ based on the publication year (4*16*3300000000=211200000000 FLOP/s per machine) Compute: 4*211200000000*168*60*60*0.3 = 1.53e17 Alternatively, the training set is "several hundred hours of speech", with inputs consisting of 26 frames, each frame is 10ms apart. If we assume 400h of training data, 400h/10ms/26= 5,538,461 inputs 6 * 101706240 * 5,538,461 = 3.38e15 FLOPs per e…

How it was established
Hardware

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
168 hours (7 days)

"The results we report are obtained by training for one week."

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
Training cost,SOTA improvement

I don't see any standard benchmarks that they used for evaluations

Record confidence
Likely

Sources

Where this record came from and when it was last checked.

Reference
On rectified linear units for speech processing
Last updated
28 November 2025

What the numbers mean

About this model

ReLU-Speech was published by Google,University of Toronto,New York University (NYU), in the country recorded as United States of America, during May 2013. The publishing organisation is categorised as industry,Academia,Academia.

It works in the domain of Speech, and is recorded as performing the task of speech recognition (ASR).

Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.

What went into building it

Producing it required arithmetic totalling around 1.3 × 10¹⁷ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.

It is tracked in the underlying dataset for one reason in particular: training cost,SOTA improvement.

Answers

ReLU-Speech — common questions

01

ReLU-Speech— how many parameters does it have?

It has a parameter count of 101.7M. "The overall input dimensionality is 1040," "All layers of our networks have 2560 hidden units " "used to generate 7969 context-dependent tied acoustic states" Largest model: 12 hidden layers (Fig 4) Parameters: 1040*2560+12*2560*2560+2560*7969=101706240. 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.

02

ReLU-Speech— who created it?

It was published by Google,University of Toronto,New York University (NYU), based in United States of America, an organisation categorised as industry,Academia,Academia.

03

ReLU-Speech— 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.

04

ReLU-Speech— what is it used for?

It works in the domain of Speech, and is recorded as handling the task of speech recognition (ASR). Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

05

ReLU-Speech— how much compute was used to train it?

Training consumed around 1.3 × 10¹⁷ 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.

06

ReLU-Speech— 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.

07

ReLU-Speech— 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.

Source

Original publication

Record last updated 28 November 2025

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.