LSTM LM

Closed weights RWTH Aachen University 102.7M parameters September 2012

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
RWTH Aachen University
Organisation type
Academia
Country
Germany
Published
9 September 2012
Authors
M. Sundermeyer, R. Schlüter, H. Ney

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
102.7M

Multiple models were trained, the largest on transcribed French podcast data. "We trained an LSTM LM using 300 hidden nodes and 27 M running words of indomain training data" "Corpus sizes in number of running words; the vocabulary size of the Treebank corpus is 10 K, for Quaero French it is 170 K" Embedding and unembedding: 2*170000*300=102000000 LSTM: 4*600*300=720000 Total: 102000000+720000=102720000=1.03e8 (Assuming the embedding dimension is the same as the LSTM layer)

Training data
27,000,000 tokens

"We trained an LSTM LM using 300 hidden nodes and 27 M running words of indomain training data."

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
1.7 × 10¹⁶ FLOP

FLOP per input for LSTM layer is 4*2*(M+N)*M, for N inputs and M outputs. Embedding FLOPs: 2 * 170000 * 300 = 102,000,000 LSTM FLOPs: 4 * 2 * (300 + 300) * 300 = 1,440,000 Unembedding FLOPs: 2 * 170000 * 300 = 102,000,000 Total: 205,440,000 FLOPs per word per forward pass For 27M training input words and including backward passes: 27M * 3 * 205,440,000 = 1.66e16 However, it sounds like they're doing something with a secondary acoustic model, so this may be an underestimate.

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.

Why it is tracked
Highly cited
Record confidence
Speculative

Sources

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

Reference
LSTM Neural Networks for Language Modeling
Last updated
28 November 2025

What the numbers mean

About this model

LSTM LM was published by RWTH Aachen University, in Germany, in September 2012. The organisation is categorised as 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.

How it was trained

Producing it required around 1.7 × 10¹⁶ FLOP of arithmetic, which is a statement about the training budget rather than about inference.

The training set ran to roughly 27,000,000 tokens.

It is tracked in the underlying dataset for one reason in particular: highly cited.

Answers

LSTM LM — common questions

01

Is LSTM LM open source?

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

02

How many parameters does LSTM LM have?

LSTM LM has 102.7M parameters. Multiple models were trained, the largest on transcribed French podcast data. "We trained an LSTM LM using 300 hidden nodes and 27 M running words of indomain training data" "Corpus sizes in number of running words; the vocabulary size of the Treebank corpus is 10 K, for Quaero French it is 170 K" Embedding and unembedding: 2*170000*300=102000000 LSTM: 4*600*300=720000 Total: 102000000+720000=102720000=1.03e8 (Assuming the embedding dimension is the same as the LSTM layer). 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

Who created LSTM LM?

LSTM LM was published by RWTH Aachen University, based in Germany, categorised as academia.

04

When was LSTM LM released?

LSTM LM was published in September 2012. 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

What is LSTM LM used for?

LSTM LM works in Language, and is recorded as handling language modeling. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

06

How much compute was used to train LSTM LM?

Around 1.7 × 10¹⁶ 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.

07

What GPU do I need to run LSTM LM?

None. LSTM 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.

Source

Original publication

Record last updated 28 November 2025

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.