DeBERTa TPS calculator

Open weights Microsoft 1.5B parameters June 2021

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

818 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Tesla C1080

4 GB · Q8_0 · 24.6 tok/s

Fastest card

B200

2,259 tok/s · 180 GB

Which GPUs can run DeBERTa?

Set the inputs, read the answer

Context length tokens

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

Minimum quality

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

Search cards

Card memory

Also show cards that cannot run it

818 cards match

Calculating

					Needs	Quantisation	Fit
2,259 tok/s 1,355–3,614 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	2.3 GB	Q8_0	Comfortable
2,259 tok/s 1,355–3,614 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	2.3 GB	Q8_0	Comfortable
1,804 tok/s 1,082–2,886 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	2.3 GB	Q8_0	Comfortable
1,804 tok/s 1,082–2,886 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	2.3 GB	Q8_0	Comfortable
1,443 tok/s 866–2,308 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	2.3 GB	Q8_0	Comfortable
1,381 tok/s 828–2,209 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	2.3 GB	Q8_0	Comfortable
1,381 tok/s 828–2,209 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	2.3 GB	Q8_0	Comfortable
1,321 tok/s 793–2,114 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	2.3 GB	Q8_0	Comfortable
1,173 tok/s 704–1,876 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	2.3 GB	Q8_0	Comfortable
1,173 tok/s 704–1,876 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	2.3 GB	Q8_0	Comfortable
1,173 tok/s 704–1,876 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	2.3 GB	Q8_0	Comfortable
1,112 tok/s 667–1,780 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable
949 tok/s 569–1,518 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable
949 tok/s 569–1,518 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	2.3 GB	Q8_0	Comfortable
949 tok/s 569–1,518 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable
949 tok/s 569–1,518 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable
949 tok/s 569–1,518 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable
722 tok/s 433–1,156 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	2.3 GB	Q8_0	Comfortable
722 tok/s 433–1,156 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	2.3 GB	Q8_0	Comfortable
602 tok/s 361–963 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	2.3 GB	Q8_0	Comfortable
589 tok/s 353–943 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	2.3 GB	Q8_0	Comfortable
576 tok/s 346–922 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	2.3 GB	Q8_0	Comfortable
576 tok/s 346–922 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	2.3 GB	Q8_0	Comfortable
576 tok/s 346–922 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	2.3 GB	Q8_0	Comfortable
576 tok/s 346–922 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	2.3 GB	Q8_0	Comfortable

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: Microsoft
Organisation type: Industry
Country: United States of America
Published: 10 June 2021
Authors: Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen

What it does

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

Domain: Language
Task: Language modeling/generation, Question answering
Approach: Self-supervised learning
Numerical format: FP16

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: 1.5B
Training data: 20,800,000,000 tokens
Epochs: 49.2

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: 2.6 × 10²² FLOP
How it was established: Hardware

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 V100
Chips used: 256
Wall-clock time: 720 hours (30 days)
Power draw: 155.4 kW
Compute cost: $6,682

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

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,SOTA improvement
Record confidence: Confident
Citations: 3,764

Sources

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

Reference: DeBERTa: Decoding-enhanced BERT with Disentangled Attention
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run DeBERTa

Ranked by estimated tokens per second, newest card first where speeds tie. Because generation is bound by memory bandwidth, this ordering follows bandwidth rather than any gaming benchmark.

The smallest GPUs that still run DeBERTa

The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.

What the numbers mean

The hardware side

Minimum card

Tesla C1080

Memory needed

2.3 GB

Fastest

2,259 tok/s

DeBERTa is small enough at 1.5B parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

At the low end, a Tesla C1080 handles it — 4 GB, at Q8_0, for about 24.6 tokens per second.

The quickest result comes from a B200 at around 2,259 tokens per second — its 8,000 GB/s of bandwidth is what buys that.

What this model is

DeBERTa was published by Microsoft, in United States of America, in June 2021. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling/generation, Question answering.

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

What decides the speed

Half the cards that hold it manage more than 63.4 tokens per second, and 796 exceed reading speed outright.

Every weight participates in every token here, so bandwidth is the whole story: the ranking below is effectively a ranking of memory throughput.

Memory here is estimated from size rather than computed from the architecture, which is not recorded for this model — the numbers are indicative rather than exact.

Training and provenance

Producing it required around 2.6 × 10²² FLOP of arithmetic, on NVIDIA V100, which is a statement about the training budget rather than about inference.

Around 20,800,000,000 tokens went into training it.

Its inclusion criterion is highly cited,SOTA improvement.

Step by step

How to choose a GPU for DeBERTa

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

01

Read the memory figure first

Every card here has been checked against DeBERTa — around 2.3 GB at Q8_0. Capacity is the gate — a card either holds it or it does not.
02

Set the context length you will work at

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for DeBERTa.
03

Choose how far you will compress it

Each card runs the least-compressed copy it can hold — Q8_0 on the smallest card that fits. Setting a floor drops the cards that only manage DeBERTa by squeezing it further than you would want.
04

Rank by throughput rather than spec sheet

Sort by speed to see how cards rank for DeBERTa. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 2,259 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage DeBERTa from those with room to spare. Buy for the second if the context might grow.
06

Open the card you have settled on

Following a card through to its own page shows every other model it can hold, which is the question that follows once DeBERTa is settled.

Answers

DeBERTa — common questions

How many parameters does DeBERTa have?

DeBERTa has 1.5B parameters. "...we scale up DeBERTa by training a larger version that consists of 48 Transform layers with 1.5 billion parameters" Other versions are smaller and use a smaller pre-training dataset. These are distinguished in the paper (e.g. DeBERTa1.5B is the version of DeBERTa with 1.5 billion 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.

Who created DeBERTa?

DeBERTa was published by Microsoft, based in United States of America, categorised as industry.

When was DeBERTa released?

DeBERTa was published in June 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

What is DeBERTa used for?

DeBERTa works in Language, and is recorded as handling language modeling/generation, Question answering. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

Where can I download DeBERTa?

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

How much compute was used to train DeBERTa?

Around 2.6 × 10²² FLOP, on NVIDIA V100. 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.

Can I run DeBERTa if it does not fit in my GPU?

Partly. Layers that do not fit sit in system memory and run at a fraction of the speed, so a mostly-offloaded DeBERTa is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run DeBERTa faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold DeBERTa on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for DeBERTa?

A larger card holds a more accurate copy. Across the cards that run DeBERTa, 1 compression levels are used; the floor control above pins it to one.

How accurate are these DeBERTa speed estimates?

Every figure is derived from memory bandwidth and model size, not benchmarked. That is why each is published as a range such as 1,355–3,614 tok/s on the B200 rather than a single number.

What GPU do I need to run DeBERTa?

The smallest card in our catalogue that holds DeBERTa is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 2.3 GB, and produces roughly 24.6 tokens per second. 818 cards in total can run it.

How fast is DeBERTa on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 2,259 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 796 of the cards that can run DeBERTa clear that.

How much VRAM does DeBERTa need?

About 2.3 GB at Q8_0 compression, which is what the smallest card that runs it uses. Less compression needs more: the figures in the memory column above are recalculated for each card, because each one holds the least-compressed version it can.

Can I run DeBERTa on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 2.3 GB and generating roughly 421 tokens per second — a comfortable fit.

Can I run DeBERTa on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 2.3 GB and generating roughly 258 tokens per second — a comfortable fit.

Can I run DeBERTa on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 2.3 GB and generating roughly 319 tokens per second — a comfortable fit.

Can I run DeBERTa on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 2.3 GB and generating roughly 378 tokens per second — a comfortable fit.

Is DeBERTa open source?

Its weights are published, so DeBERTa 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.

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.

All AI models