MSRA (C, PReLU)

Closed weights Microsoft Research 87M parameters February 2015

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
Microsoft Research
Organisation type
Industry
Country
United States of America
Published
6 February 2015
Authors
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun

What it does

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

Domain
Vision
Task
Image classification
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
87M

I used the architecture in table 3 I ignored biases, and assumed a SPP bin size of 256 3*7*7*96+96*3*3*384+384*3*3*384*5+384*3*3*768+768*3*3*768*5+768*3*3*896+896*3*3*896*5+896*(7*7+3*3+2*2+1)*4096+4096*4096+4096*1000=330581792

Training data
1,280,000 tokens

"We perform the experiments on the 1000-class ImageNet 2012 dataset", paper; ImageNet 2012 train set size from https://huggingface.co/datasets/imagenet-1k

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
2.4 × 10¹⁹ FLOP

"training C on eight K40 GPUs, takes about 3-4 weeks" 0.33 util rate (From Imagenet paper-data, Besiroglu et al., forthcoming)

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.

Training hardware
NVIDIA Tesla K40t
Wall-clock time
588 hours (24.5 days)
Compute cost
$1,394

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Frontier model
Yes
Why it is tracked
Highly cited
Record confidence
Confident
Citations
20,475

Sources

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

Reference
Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
Last updated
25 May 2026

What the numbers mean

What this model is

MSRA (C, PReLU) was published by Microsoft Research, in the country recorded as United States of America, during February 2015. It comes out of an organisation categorised as industry.

It works in the domain of Vision, and is recorded as performing the task of image classification.

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

Training and provenance

The training run consumed about 2.4 × 10¹⁹ FLOP, on hardware recorded as NVIDIA Tesla K40t. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 1,280,000 tokens of text.

Its inclusion criterion: highly cited.

Answers

MSRA (C, PReLU) — common questions

01

MSRA (C, PReLU)— who created it?

It was published by Microsoft Research, based in United States of America, an organisation categorised as industry.

02

MSRA (C, PReLU)— when was it released?

It was published in February 2015. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

03

MSRA (C, PReLU)— what is it used for?

It works in the domain of Vision, and is recorded as handling the task of image classification. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

04

MSRA (C, PReLU)— how much compute was used to train it?

Training consumed around 2.4 × 10¹⁹ FLOP, on hardware recorded as NVIDIA Tesla K40t. 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.

05

MSRA (C, PReLU)— 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.

06

MSRA (C, PReLU)— 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.

07

MSRA (C, PReLU)— how many parameters does it have?

It has a parameter count of 87M. I used the architecture in table 3 I ignored biases, and assumed a SPP bin size of 256 3*7*7*96+96*3*3*384+384*3*3*384*5+384*3*3*768+768*3*3*768*5+768*3*3*896+896*3*3*896*5+896*(7*7+3*3+2*2+1)*4096+4096*4096+4096*1000=330581792. 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.

Source

Original publication

Record last updated 25 May 2026

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.