YOLOv3

Closed weights University of Washington 56.9M parameters April 2018

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
University of Washington
Organisation type
Academia
Country
United States of America
Published
8 April 2018
Authors
Joseph Redmon, Ali Farhadi

What it does

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

Domain
Vision
Task
Object detection

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
56.9M

Feature extractor (ignoring biases) 32*3*3*3 + 64*3*3*32 + 32*1*1*64 + 64*3*3*32 + 128*3*3*64 + 2*(64*1*1*128 + 128*3*3*64) + 256*3*3*128 + 8*(128*1*1*256 + 256*3*3*128) + 512*3*3*256 + 8*(256*1*1*512 + 512*3*3*256) + 1024*3*3*512 + 4*(512*1*1*1024 + 1024*3*3*512) + 4*4*1024*1000 source: table 1 This is assuming the average pooling step changes the output size from 8x8 to 4x4. The weights file is 237MB. If the weights are saved as float32, 4 bytes per weight, then there are approximately 2…

Training data
5,430,000 tokens

Source: https://image-net.org/download.php

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 × 10¹⁹ FLOP

We use the formula training_compute = ops_per_forward_pass * 3.5 * n_epochs * n_examples Assuming 160 epochs of training as in https://arxiv.org/pdf/1612.08242.pdf Table 2: 18700000000 operations 18700000000 ops * 3.5 *160 epochs * 1281167 images

How it was established
Operation counting

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 M40,NVIDIA GeForce GTX TITAN X

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

code and weights, unclear license: https://pjreddie.com/darknet/yolo/

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
Record confidence
Likely
Citations
25,147

Sources

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

Reference
YOLOv3: An Incremental Improvement
Last updated
25 May 2026

What the numbers mean

What this model is

YOLOv3 was published by University of Washington, in the country recorded as United States of America, during April 2018. The publishing organisation is categorised as academia.

It works in the domain of Vision, and is recorded as performing the task of object detection.

Its weights were never published, so it can only be reached through its provider. No graphics card changes that.

Training and provenance

Training it took a computation budget of roughly 1.3 × 10¹⁹ FLOP, on hardware recorded as NVIDIA M40,NVIDIA GeForce GTX TITAN X. That figure measures what producing the model cost, and has no bearing on how fast it answers.

It was trained on a corpus of about 5,430,000 tokens of text.

Its inclusion criterion: highly cited.

Answers

YOLOv3 — common questions

01

YOLOv3— is it open source?

No. Its weights have not been published, so it exists only as a service controlled by its owner.

02

YOLOv3— how many parameters does it have?

It has a parameter count of 56.9M. Feature extractor (ignoring biases) 32*3*3*3 + 64*3*3*32 + 32*1*1*64 + 64*3*3*32 + 128*3*3*64 + 2*(64*1*1*128 + 128*3*3*64) + 256*3*3*128 + 8*(128*1*1*256 + 256*3*3*128) + 512*3*3*256 + 8*(256*1*1*512 + 512*3*3*256) + 1024*3*3*512 + 4*(512*1*1*1024 + 1024*3*3*512) + 4*4*1024*1000 source: table 1 This is assuming the average pooling step changes the output size from 8x8 to 4x4. The weights file is 237MB. If the weights are saved as float32, 4 bytes per weight, then there are approximately 237M/4=59M parameters, consistent with the calculation above. 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.

03

YOLOv3— who created it?

It was published by University of Washington, based in United States of America, an organisation categorised as academia.

04

YOLOv3— when was it released?

It was published in April 2018. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

05

YOLOv3— what is it used for?

It works in the domain of Vision, and is recorded as handling the task of object detection. 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.

06

YOLOv3— how much compute was used to train it?

Training consumed around 1.3 × 10¹⁹ FLOP, on hardware recorded as NVIDIA M40,NVIDIA GeForce GTX TITAN X. 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.

07

YOLOv3— 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.

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.