Neural LM

Closed weights University of Montreal / Université de Montréal 6.9M parameters November 2000

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 Montreal / Université de Montréal
Organisation type
Academia
Country
Canada
Published
28 November 2000
Authors
Yoshua Bengio, Réjean Ducharme, Pascal Vincent, Christian Janvin

What it does

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

Domain
Language
Task
Language modeling

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
6.9M

(30959*100) + (8*100*120) + (120*30959) = 6,906,980 "This is obtained with a network with the direct architecture, 100 randomly initialized words features, 120 hidden units, and n = 8 words of context." "The Hansard corpus (Canadian parliament proceedings, French version) is a stream of about 34 million words, of which 32 millions (set A) was used for training, 1.1 million (set B) was used for validation, and 1.2 million (set C) was used for out-of-sample tests. The original data has 106, 936 di…

Training data
32,000,000 tokens

The Hansard corpus (Canadian parliament proceedings, French version) is a stream of about 34 million words, of which 32 millions (set A) was used for training, 1.1 million (set B) was used for validation, and 1.2 million (set C) was used for out-of-sample tests.

Epochs
10

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
6.3 × 10¹⁵ FLOP

The authors use a trick to avoid having to calculate the final layer for all possible words in the vocabulary. They precompute a "short list" of the most common word following any 2 precursor words with a smoothed trigram model, and then only calculate the softmax over words on the short list. This means only a negligible fraction of the unembedding parameters get used, so the effective number of parameters appears to be (30959*100) + (8*100*120) = 3,191,900 "Apparent convergence of the stochas…

How it was established
Operation counting

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
Training cost,Historical significance,Highly cited
Record confidence
Confident
Citations
7,637

Sources

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

Reference
A Neural Probabilistic Language Model
Last updated
1 January 2026

What the numbers mean

About this model

Neural LM was published by University of Montreal / Université de Montréal, in Canada, in November 2000. It comes out of academia.

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.

What went into building it

Training it took roughly 6.3 × 10¹⁵ FLOP of computation — a measure of what producing the model cost, not of how fast it answers.

It was trained on about 32,000,000 tokens of text.

Its inclusion criterion is training cost,Historical significance,Highly cited.

Answers

Neural LM — common questions

01

What GPU do I need to run Neural LM?

None. Neural LM 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.

02

Is Neural LM open source?

The licensing for Neural LM was never recorded in our source data. We treat unstated licensing as closed, because an unrecorded licence is not one to rely on.

03

How many parameters does Neural LM have?

Neural LM has 6.9M parameters. (30959*100) + (8*100*120) + (120*30959) = 6,906,980 "This is obtained with a network with the direct architecture, 100 randomly initialized words features, 120 hidden units, and n = 8 words of context." "The Hansard corpus (Canadian parliament proceedings, French version) is a stream of about 34 million words, of which 32 millions (set A) was used for training, 1.1 million (set B) was used for validation, and 1.2 million (set C) was used for out-of-sample tests. The original data has 106, 936 different words, and those with frequency <= 10 were merged into a single token, yielding IVI = 30,959 different words.". 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.

04

Who created Neural LM?

Neural LM was published by University of Montreal / Université de Montréal, based in Canada, categorised as academia.

05

When was Neural LM released?

Neural LM was published in November 2000. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

06

What is Neural LM used for?

Neural LM works in Language, and is recorded as handling language modeling. These are the areas it was designed around; they describe intent rather than a hard boundary.

07

How much compute was used to train Neural LM?

Around 6.3 × 10¹⁵ FLOP. 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.

Source

Original publication

Record last updated 1 January 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.