Turing-NLG

Closed weights Microsoft 17B parameters February 2020

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
Microsoft
Organisation type
Industry
Country
United States of America
Published
13 February 2020
Authors
Corby Rosset

What it does

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

Domain
Language
Task
Text autocompletion, Language generation, Text summarization
Approach
Self-supervised learning

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
17B
Training data
46,400,000,000 tokens

Authors say they pretrain on the same data as for Megatron-LM. From the Megatron-LM paper: https://arxiv.org/pdf/1909.08053.pdf "The resulting aggregate corpus contains 174 GB of deduplicated text." 174GB * 2e8words/GB = 3.48e10 words 3.48e10 words (if english) *4/3 = 46400000000 tokens confidence - likely

Epochs
3.39
Batch size
524,288

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
1.6 × 10²² FLOP

source: https://lair.lighton.ai/akronomicon/ 157 PF-days * 3600 * 24 * 10^15 = 1.35648e+22 archived: https://github.com/lightonai/akronomicon/tree/main/akrodb 6ND=6*17000000000*46400000000=4.7328e+21 (confidence regarding dataset size - likely) Authors of "AI and Memory Wall" (https://github.com/amirgholami/ai_and_memory_wall) estimated model's training compute as 28,000,000 PFLOP = 2.8*10^22 FLOP

How it was established
Third-party estimation,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 Tesla V100 DGXS 32 GB
Chips used
256
Power draw
130.9 kW
Compute cost
$51,660

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
Unreleased

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

from paper: "Turing Natural Language Generation (T-NLG) is a 17 billion parameter language model by Microsoft that outperforms the state of the art on many downstream NLP tasks"

Record confidence
Likely
Citations
114
Benchmark data
Turing-NLG

Sources

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

Reference
Turing-NLG: A 17-billion-parameter language model by Microsoft
Last updated
28 November 2025

What the numbers mean

What this model is

Turing-NLG was published by Microsoft, in United States of America, in February 2020. It comes out of industry.

It works in Language, and is recorded as doing text autocompletion, Language generation, Text summarization.

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

How it was trained

Producing it required around 1.6 × 10²² FLOP of arithmetic, on NVIDIA Tesla V100 DGXS 32 GB, which is a statement about the training budget rather than about inference.

It was trained on about 46,400,000,000 tokens of text.

The reason it appears in this catalogue at all is sOTA improvement.

Answers

Turing-NLG — common questions

01

How many parameters does Turing-NLG have?

Turing-NLG has 17B parameters. 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.

02

Who created Turing-NLG?

Turing-NLG was published by Microsoft, based in United States of America, categorised as industry.

03

When was Turing-NLG released?

Turing-NLG was published in February 2020. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

04

What is Turing-NLG used for?

Turing-NLG works in Language, and is recorded as handling text autocompletion, Language generation, Text summarization. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

05

How much compute was used to train Turing-NLG?

Around 1.6 × 10²² FLOP, on NVIDIA Tesla V100 DGXS 32 GB. 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.

06

What GPU do I need to run Turing-NLG?

None. Turing-NLG 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.

07

Is Turing-NLG open source?

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

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.