DARTS

Closed weights DeepMind,Carnegie Mellon University (CMU) 33M 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
DeepMind,Carnegie Mellon University (CMU)
Organisation type
Industry,Academia
Country
United Kingdom of Great Britain and Northern Ireland, United States of America
Published
24 June 2018
Authors
Hanxiao Liu, Karen Simonyan, Yiming Yang

What it does

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

Domain
Language
Task
Language modeling, Neural Architecture Search - NAS

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
33M

33M (Table 4) - parameters reported for PTB, but they say it is the same for WT-2 " WIKITEXT-2 We use embedding and hidden sizes 700, weight decay 5×10−7, and hidden-node variational dropout 0.15. Other hyperparameters remain the same as in our PTB experiments."

Training data
2,000,000 tokens

300 epochs for PTB (supposedly the same for WT-2

Epochs
300

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
3.2 × 10¹⁷ FLOP

6 FLOP / parameter / token * 33000000 parameters * 2000000 tokens * 300 epochs = 1.188e+17 FLOP 11340000000000 FLOP / second / GPU * 1 GPU * 72 hours * 3600 sec / hour * 0.3 [assumed utilization] = 8.817984e+17 FLOP sqrt(1.188e+17*8.817984e+17) = 3.2366286e+17 FLOP 'Speculative' confidence since many variables are assumed based on PTB model training

How it was established
Operation counting,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 GeForce GTX 1080 Ti
Chips used
1
Wall-clock time
72 hours

for PTB (they say WT-2 training is similar): The training takes 3 days on a single 1080Ti GPU with our PyTorch implementation 3*24 = 72 hours

Power draw
285 W

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
Open source

apache 2, training/test for wikitext: https://github.com/quark0/darts/tree/master/rnn

How it is classified

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

Record confidence
Speculative
Citations
4,929
Benchmark data
DARTS

Sources

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

Reference
DARTS: Differentiable Architecture Search
Last updated
25 May 2026

What the numbers mean

About this model

DARTS was published by DeepMind,Carnegie Mellon University (CMU), in United Kingdom of Great Britain and Northern Ireland, in June 2018. The organisation is categorised as industry,Academia.

It works in Language, and is recorded as doing language modeling, Neural Architecture Search - NAS.

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

Training and provenance

The training run consumed about 3.2 × 10¹⁷ FLOP, on NVIDIA GeForce GTX 1080 Ti. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

It was trained on about 2,000,000 tokens of text.

Answers

DARTS — common questions

01

What is DARTS used for?

DARTS works in Language, and is recorded as handling language modeling, Neural Architecture Search - NAS. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

02

How much compute was used to train DARTS?

Around 3.2 × 10¹⁷ FLOP, on NVIDIA GeForce GTX 1080 Ti. 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.

03

What GPU do I need to run DARTS?

None. DARTS 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.

04

Is DARTS open source?

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

05

How many parameters does DARTS have?

DARTS has 33M parameters. 33M (Table 4) - parameters reported for PTB, but they say it is the same for WT-2 " WIKITEXT-2 We use embedding and hidden sizes 700, weight decay 5×10−7, and hidden-node variational dropout 0.15. Other hyperparameters remain the same as in our PTB experiments.". 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.

06

Who created DARTS?

DARTS was published by DeepMind,Carnegie Mellon University (CMU), based in United Kingdom of Great Britain and Northern Ireland, categorised as industry,Academia.

07

When was DARTS released?

DARTS 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.

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.