4 layer QRNN (h=2500)

Closed weights Salesforce Research 151M parameters March 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
Salesforce Research
Organisation type
Industry
Country
United States of America
Published
22 March 2018
Authors
Stephen Merity, Nitish Shirish Keskar, Richard Socher

What it does

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

Domain
Language
Task
Language modeling
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
151M

Table 6

Training data
103,000,000 tokens

"The model was trained for 12 hours (14 epochs)"

Epochs
14

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
5.9 × 10¹⁷ FLOP

20670000000000 FLOP / sec / GPU [fp16 assumed] * 1 GPU * 12 hours * 3600 sec / hour * 0.3 [assumed utilization] = 2.678832e+17FLOP 6 FLOP / token / parameter * 151000000 parameters * 103000000 tokens * 14 epochs = 1.306452e+18 FLOP sqrt(2.678832e+17*1.306452e+18) = 5.9158815e+17 FLOP __________________ in the algorithmic progress paper the estimation was 2.4 × 10^17 FLOP under assumption of 26M parameters

How it was established
Hardware,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 Quadro GP100
Chips used
1
Chip-hours
12
Wall-clock time
12 hours

"Results are obtained in only 12 hours (WikiText-103) to 2 days (enwik8) using a single modern GPU"

Power draw
268 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

BSD-3 license: https://github.com/salesforce/awd-lstm-lm

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
SOTA improvement

"QRNNs achieve stateof-the-art results on character-level (Penn Treebank, enwik8) and word-level (WikiText-103) datasets, respectively"

Record confidence
Likely
Citations
183
Benchmark data
4 layer QRNN (h=2500)

Sources

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

Reference
An Analysis of Neural Language Modeling at Multiple Scales
Last updated
11 February 2026

What the numbers mean

What this model is

4 layer QRNN (h=2500) was published by Salesforce Research, in United States of America, in March 2018. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling.

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

How it was trained

Producing it required around 5.9 × 10¹⁷ FLOP of arithmetic, on NVIDIA Quadro GP100, which is a statement about the training budget rather than about inference.

The training set ran to roughly 103,000,000 tokens.

It is tracked in the underlying dataset for one reason in particular: sOTA improvement.

Answers

4 layer QRNN (h=2500) — common questions

01

What is 4 layer QRNN (h=2500) used for?

4 layer QRNN (h=2500) works in Language, and is recorded as handling language modeling. 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.

02

How much compute was used to train 4 layer QRNN (h=2500)?

Around 5.9 × 10¹⁷ FLOP, on NVIDIA Quadro GP100. 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 4 layer QRNN (h=2500)?

None. 4 layer QRNN (h=2500) 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 4 layer QRNN (h=2500) open source?

No. 4 layer QRNN (h=2500) has not had its weights published, so it exists only as a service controlled by its owner.

05

How many parameters does 4 layer QRNN (h=2500) have?

4 layer QRNN (h=2500) has 151M parameters. Table 6. 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 4 layer QRNN (h=2500)?

4 layer QRNN (h=2500) was published by Salesforce Research, based in United States of America, categorised as industry.

07

When was 4 layer QRNN (h=2500) released?

4 layer QRNN (h=2500) was published in March 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 11 February 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.