QT-Opt

Closed weights Google Brain,University of California (UC) Berkeley 1.2M parameters June 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
Google Brain,University of California (UC) Berkeley
Organisation type
Industry,Academia
Country
United States of America
Published
27 June 2018
Authors
Dmitry Kalashnikov, Alex Irpan, Peter Pastor, Julian Ibarz, Alexander Herzog, Eric Jang, Deirdre Quillen, Ethan Holly, Mrinal Kalakrishnan, Vincent Vanhoucke, Sergey Levine

What it does

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

Domain
Robotics, Vision
Task
Robotic manipulation
Approach
Reinforcement learning

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
1.2M

"The Q-function Qθ(s, a) is represented in our system by a large convolutional neural network with 1.2M parameters"

Training data
11,600,000 tokens

Observations take up 4TB of disk space, and the input space is a 472x472 RGB image. Assuming 24 bit depth color (8 bits per channel), that suggests 472 * 472 * 3 * 8 bits = 668.352 kB per image (this could be off by a factor of 2 depending on actual bit depth) 4 TB / 668.352 kB = 5,984,870 images; around 10 per grasp attempt. 15M gradient steps with batchsize 32 implies: 15M steps * 32 images/step * 1/5984870 images ~= each image seen 80 times

Epochs
80

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
1.4 × 10¹⁹ FLOP

"We distribute training across 10 GPUs, using asynchronous SGD with momentum... This system allows us to train the Q-function at 40 steps per second with a batch size of 32 across 10 NVIDIA P100 GPUs." "We found empirically that a large number of gradient steps (up to 15M) were needed to train an effective Q-function..." 15M steps * 0.025 seconds/step * 9.30E+12 FLOP/sec/GPU * 0.4 utilization * 10 GPU = 1.395E+19

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 P100
Wall-clock time
104 hours

"We distribute training across 10 GPUs, using asynchronous SGD with momentum... This system allows us to train the Q-function at 40 steps per second with a batch size of 32 across 10 NVIDIA P100 GPUs." "We found empirically that a large number of gradient steps (up to 15M) were needed to train an effective Q-function..." 15M steps * 0.025 seconds/step * 1/3600 hours/second = 104.2 hours

Compute cost
$1,318

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

How it is classified

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

Record confidence
Likely
Citations
1,692

Sources

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

Reference
QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation
Last updated
25 May 2026

What the numbers mean

About this model

QT-Opt was published by Google Brain,University of California (UC) Berkeley, in United States of America, in June 2018. The organisation is categorised as industry,Academia.

It works in Robotics, Vision, and is recorded as doing robotic manipulation.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

Training and provenance

The training run consumed about 1.4 × 10¹⁹ FLOP, on NVIDIA P100. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

Around 11,600,000 tokens went into training it.

Answers

QT-Opt — common questions

01

How many parameters does QT-Opt have?

QT-Opt has 1.2M parameters. "The Q-function Qθ(s, a) is represented in our system by a large convolutional neural network with 1.2M parameters". 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

Who created QT-Opt?

QT-Opt was published by Google Brain,University of California (UC) Berkeley, based in United States of America, categorised as industry,Academia.

03

When was QT-Opt released?

QT-Opt was published in June 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.

04

What is QT-Opt used for?

QT-Opt works in Robotics, Vision, and is recorded as handling robotic manipulation. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

05

How much compute was used to train QT-Opt?

Around 1.4 × 10¹⁹ 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.

06

What GPU do I need to run QT-Opt?

None. QT-Opt 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

Is QT-Opt open source?

No. QT-Opt has not had its weights published, so it exists only as a service controlled by its owner.

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.