QMoE: compressed SwitchTransformer TPS calculator

Open weights Institute of Science and Technology Austria (ISTA),Neural Magic 1.6T parameters October 2023

Each card below is assessed against this model at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from the card's memory bandwidth and the size of the model once compressed.

Calculated for this model

0 cards that can run it

818 cards we hold specifications for

Which GPUs can run QMoE: compressed SwitchTransformer?

Set the inputs, read the answer

A longer conversation needs more memory, which can push this model off smaller cards.

Hides cards that would only fit the model by compressing it below this point.

0 cards match

Calculating
Needs Quantisation Fit

No card in our catalogue can run this model with these settings.

Speeds are estimates for a single request — one conversation at a time — calculated from memory bandwidth, model size and quantisation. Real throughput varies with the inference runtime and its version. Figures published by hardware vendors measure many simultaneous requests and are much higher.

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
Institute of Science and Technology Austria (ISTA),Neural Magic
Organisation type
Academia,Industry
Country
Austria, United States of America
Published
25 October 2023
Authors
Elias Frantar, Dan Alistarh

What it does

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

Domain
Language
Task
Language modeling
Base model
Switch

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
1.6T

"Concretely, QMoE can compress the 1.6 trillion parameter SwitchTransformer-c2048 model to less than 160GB" Same parameter count as base model. This paper compresses the model; there is no learning from data.

Training data
576,000,000,000 tokens

In this paper, the authors compress the SwitchTransformer-c2048 model. There is no training dataset used.

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.

Fine-tuning compute
1 × 10¹⁸ FLOP

(1) * (38.71 * 10 ** 12) * (0.3) * (24 * 3600) = 1003363200000000000 (num gpu) * (peak flop) * (assumed utilization rate) * (time in seconds) from the paper: "This allows us to apply data-dependent compression to massive MoEs, while preserving the key feature of post-training compression techniques: the ability to perform effective compression using only modest computational resources, e.g., a single NVIDIA A6000 GPU and less than one day of compute." A6000 have 38.71 TFLOPs from https://www.tec…

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 RTX A6000

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
Open — downloadable
Model access
Open weights (unrestricted)
Training code
Open source

apache 2.0 github.com/IST-DASLab/qmoe

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence
Confident
Citations
47

Sources

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

Reference
QMoE: Practical Sub-1-Bit Compression of Trillion-Parameter Models
Last updated
25 May 2026

What the numbers mean

The hardware side

QMoE: compressed SwitchTransformer reaches a parameter count of 1.6T. That is beyond what any single graphics card holds. Running it means either splitting it across several cards or renting hardware built for the job, and every card able to hold it alone is a datacentre part. The number that can: 0.

Background

QMoE: compressed SwitchTransformer was published by Institute of Science and Technology Austria (ISTA),Neural Magic, in the country recorded as Austria, during October 2023. The category the publisher falls under is academia,Industry.

It works in the domain of Language, and is recorded as performing the task of language modeling.

Rather than being trained from scratch, it is derived from Switch. Most models at this scale are adapted from an existing base rather than built from nothing.

The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached.

What went into building it

It was trained on a corpus of about 576,000,000,000 tokens of text.

Step by step

How to choose a GPU for QMoE: compressed SwitchTransformer

The table above has already assessed every card we hold specifications for against this model. Getting to your answer takes six steps.

  1. 01

    Read the memory figure first

    The table lists every card able to hold QMoE: compressed SwitchTransformer. No amount of processing power compensates for a card that cannot hold it.

  2. 02

    Decide how long your conversations run

    The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect, because at long context a card that handles short questions easily can be dropped by QMoE: compressed SwitchTransformer.

  3. 03

    Set a quality floor

    Each card runs the least-compressed copy it can hold. Setting a minimum quality drops the cards that only manage it by squeezing further than you would want, and holds the comparison at one level.

  4. 04

    Sort by speed

    The speed ordering is effectively an ordering by memory bandwidth, for QMoE: compressed SwitchTransformer. It will not match a gaming ordering, because generation is bound by memory bandwidth.

  5. 05

    Check the fit verdict before buying

    Tight means it loads and works with no room to raise the context later, in the case of QMoE: compressed SwitchTransformer. Comfortable means you can grow the context later. That difference matters more than a few tokens per second, so buy for comfortable if you expect to.

  6. 06

    See what else that card runs

    Following a card through to its own page shows every other model it can hold, which is the question that follows once you have settled on QMoE: compressed SwitchTransformer.

Answers

QMoE: compressed SwitchTransformer — common questions

01

QMoE: compressed SwitchTransformer— what is it used for?

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

02

QMoE: compressed SwitchTransformer— where can I download it?

The weights are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

03

QMoE: compressed SwitchTransformer— can I run it if it does not fit in my GPU?

It can be split between the card and system memory, but it generates painfully slowly that way. The nearest miss we calculate falls short by 691.9 GB. Every figure here assumes the whole model is resident on the card.

04

QMoE: compressed SwitchTransformer— would two GPUs run it faster?

Two cards buy memory rather than speed, which matters only if one card cannot hold it. The number that can: 0. So a second card is rarely the answer here.

05

QMoE: compressed SwitchTransformer— why does the quantisation differ between cards?

A larger card holds a more accurate copy. The number of compression levels used across the cards that run it: 1. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

06

QMoE: compressed SwitchTransformer— how accurate are these speed estimates?

Every figure is derived from memory bandwidth and model size, not benchmarked, which is why each is published as a range rather than a single number. One example: the range beneath each figure. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

07

QMoE: compressed SwitchTransformer— is it open source?

Its weights are published, so it can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.

08

QMoE: compressed SwitchTransformer— how many parameters does it have?

It has a parameter count of 1.6T. "Concretely, QMoE can compress the 1.6 trillion parameter SwitchTransformer-c2048 model to less than 160GB" Same parameter count as base model. This paper compresses the model; there is no learning from data. 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.

09

QMoE: compressed SwitchTransformer— who created it?

It was published by Institute of Science and Technology Austria (ISTA),Neural Magic, based in Austria, an organisation categorised as academia,Industry.

10

QMoE: compressed SwitchTransformer— when was it released?

It was published in October 2023. 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 25 May 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.