RNN 1000/5 + RT09 LM (NIST RT05)

Closed weights Brno University of Technology,Johns Hopkins University 77M parameters September 2010

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
Brno University of Technology,Johns Hopkins University
Organisation type
Academia,Academia
Country
Czechia, United States of America
Published
26 September 2010
Authors
T. Mikolov, M. Karafiat, L. Burget, J. Cernocky, and S. Khudanpur

What it does

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

Domain
Speech
Task
Speech recognition (ASR), Transcription
Approach
Supervised

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
77M

"The acoustic HMMs are based on cross-word tied-states triphones trained discriminatively using MPE criteria. Feature extraction use 13 Mel-PLP’s features with deltas, double and triple deltas reduced by HLDA to 39-dimension feature vector" RNN 1000/5 indicates that the model has a 1000 units in the hidden layer and merges words that occur less than 5 times in the training set. Section 4.2 in this paper appears to indicate that NIST RT05 has a vocabulary of 50k words: https://www.fit.vut.cz/pe…

Training data
5,400,000 tokens

"Table 4: Comparison of very large back-off LMs and RNN LMs trained only on limited in-domain data (5.4M words)."

Epochs
20

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 × 10¹⁶ FLOP

"Convergence is usually achieved after 10-20 epochs." Assuming a backward-forward ratio of 2:1, since this is a shallow network 6 * 77039000 * 5.4M * 20 = 5.00e16

How it was established
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.

Wall-clock time
1,200 hours (50 days)

The biggest individual models evaluated on WSJ (probably RNN 250/2 at 27M parameters) took about 2.1e16 FLOPs to train, and training took "several weeks". Assuming a linear scaleup and several = about 3: 3 weeks * 7 days/week * 24 hr/day * 5e16/2.1e16 = 1,200 hours

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.

Frontier model
Yes
Why it is tracked
Highly cited
Record confidence
Likely
Citations
6,038

Sources

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

Reference
Recurrent neural network based language model
Last updated
11 February 2026

What the numbers mean

Background

RNN 1000/5 + RT09 LM (NIST RT05) was published by Brno University of Technology,Johns Hopkins University, in Czechia, in September 2010. academia,Academia is the category the publisher falls under.

It works in Speech, and is recorded as doing speech recognition (ASR), Transcription.

Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.

Training and provenance

Producing it required around 5 × 10¹⁶ FLOP of arithmetic, which is a statement about the training budget rather than about inference.

Around 5,400,000 tokens went into training it.

Its inclusion criterion is highly cited.

Answers

RNN 1000/5 + RT09 LM (NIST RT05) — common questions

01

What GPU do I need to run RNN 1000/5 + RT09 LM (NIST RT05)?

None. RNN 1000/5 + RT09 LM (NIST RT05) 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 RNN 1000/5 + RT09 LM (NIST RT05) open source?

No. RNN 1000/5 + RT09 LM (NIST RT05) has not had its weights published, so it exists only as a service controlled by its owner.

03

How many parameters does RNN 1000/5 + RT09 LM (NIST RT05) have?

RNN 1000/5 + RT09 LM (NIST RT05) has 77M parameters. "The acoustic HMMs are based on cross-word tied-states triphones trained discriminatively using MPE criteria. Feature extraction use 13 Mel-PLP’s features with deltas, double and triple deltas reduced by HLDA to 39-dimension feature vector" RNN 1000/5 indicates that the model has a 1000 units in the hidden layer and merges words that occur less than 5 times in the training set. Section 4.2 in this paper appears to indicate that NIST RT05 has a vocabulary of 50k words: https://www.fit.vut.cz/person/imikolov/public/rnnlm/char.pdf Using this dataset of english word frequencies ... https://www.kaggle.com/datasets/rtatman/english-word-frequency ... we can estimate that the number of words in the final vocabulary would have been around 38k when merging with threshold 5: https://colab.research.google.com/drive/1K5qH0EqXtFwTLESNtp4oelCM28GpGXt6?usp=sharing Thus the model in question would have: (39 + 1000 + 38k) * 1000 + 1000 * 38k = 77,039,000 parameters The RT09 LM they interpolate with appears to be a 4-gram model. In one sense, n-gram models have parameters roughly equal to the number of unique n-grams in the training data. However, these aren't parameters in the same sense as neural networks as they are accessed as a lookup table, so on each forward pass only one parameter is active. I choose not to include these parameters in our count. 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 RNN 1000/5 + RT09 LM (NIST RT05)?

RNN 1000/5 + RT09 LM (NIST RT05) was published by Brno University of Technology,Johns Hopkins University, based in Czechia, categorised as academia,Academia.

05

When was RNN 1000/5 + RT09 LM (NIST RT05) released?

RNN 1000/5 + RT09 LM (NIST RT05) was published in September 2010. 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 RNN 1000/5 + RT09 LM (NIST RT05) used for?

RNN 1000/5 + RT09 LM (NIST RT05) works in Speech, and is recorded as handling speech recognition (ASR), Transcription. These are the areas it was designed around; they describe intent rather than a hard boundary.

07

How much compute was used to train RNN 1000/5 + RT09 LM (NIST RT05)?

Around 5 × 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.

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.