XLM-RoBERTa TPS calculator

Open weights Facebook AI 550M parameters November 2019

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 · 67.0 tok/s

Fastest card

B200

6,160 tok/s · 180 GB

Which GPUs can run XLM-RoBERTa?

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
6,160 tok/s 3,696–9,857 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.3 GB	Q8_0	Comfortable
6,160 tok/s 3,696–9,857 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.3 GB	Q8_0	Comfortable
4,919 tok/s 2,952–7,871 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.3 GB	Q8_0	Comfortable
4,919 tok/s 2,952–7,871 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.3 GB	Q8_0	Comfortable
3,934 tok/s 2,361–6,295 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.3 GB	Q8_0	Comfortable
3,766 tok/s 2,259–6,025 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.3 GB	Q8_0	Comfortable
3,766 tok/s 2,259–6,025 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.3 GB	Q8_0	Comfortable
3,604 tok/s 2,162–5,766 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.3 GB	Q8_0	Comfortable
3,198 tok/s 1,919–5,117 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.3 GB	Q8_0	Comfortable
3,198 tok/s 1,919–5,117 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.3 GB	Q8_0	Comfortable
3,198 tok/s 1,919–5,117 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.3 GB	Q8_0	Comfortable
3,034 tok/s 1,820–4,854 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.3 GB	Q8_0	Comfortable
2,587 tok/s 1,552–4,140 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.3 GB	Q8_0	Comfortable
2,587 tok/s 1,552–4,140 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.3 GB	Q8_0	Comfortable
2,587 tok/s 1,552–4,140 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.3 GB	Q8_0	Comfortable
2,587 tok/s 1,552–4,140 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.3 GB	Q8_0	Comfortable
2,587 tok/s 1,552–4,140 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.3 GB	Q8_0	Comfortable
1,970 tok/s 1,182–3,152 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.3 GB	Q8_0	Comfortable
1,970 tok/s 1,182–3,152 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.3 GB	Q8_0	Comfortable
1,642 tok/s 985–2,627 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.3 GB	Q8_0	Comfortable
1,607 tok/s 964–2,571 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.3 GB	Q8_0	Comfortable
1,571 tok/s 943–2,513 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.3 GB	Q8_0	Comfortable
1,571 tok/s 943–2,513 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.3 GB	Q8_0	Comfortable
1,571 tok/s 943–2,513 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.3 GB	Q8_0	Comfortable
1,571 tok/s 943–2,513 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	1.3 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: Facebook AI
Organisation type: Industry
Country: United States of America
Published: 5 November 2019
Authors: Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, Veselin Stoyanov

What it does

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

Domain: Language
Task: Named entity recognition (NER), Question answering, Text classification

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: 550M
Training data: 167,000,000,000 tokens

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: 2.1 × 10²² FLOP
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.

Training hardware: NVIDIA Tesla V100 DGXS 32 GB
Chips used: 500
Power draw: 256.2 kW
Compute cost: $77,861

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 (non-commercial)
Training code: Open (non-commercial)

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,SOTA improvement
Record confidence: Confident
Citations: 8,373

Sources

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

Reference: Unsupervised Cross-lingual Representation Learning at Scale
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run XLM-RoBERTa

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 XLM-RoBERTa

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

What you need to run it

Minimum card

Tesla C1080

Memory needed

1.3 GB

Fastest

6,160 tok/s

XLM-RoBERTa reaches a parameter count of 550M. That is small enough that hardware is rarely the obstacle, including on cards several years old. The number of cards we track that can run it: 818.

The least hardware that works is Tesla C1080, with a memory capacity of 4 GB, running it at a compression of Q8_0 and producing around 67.0 tokens per second.

The quickest result comes from B200, generating roughly 6,160 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

Background

XLM-RoBERTa was published by Facebook AI, in the country recorded as United States of America, during November 2019. The publishing organisation is categorised as industry.

It works in the domain of Language, and is recorded as performing the task of named entity recognition (NER), Question answering, Text classification.

The weights being open is what puts this page in the calculator rather than only in the catalogue: it is a model you can actually hold.

Reading the throughput figures

Half the cards that hold it manage more than 173.0 tokens per second. Producing text faster than most people read it: 809 of them.

It is a dense model, so every parameter is read for every token produced. That makes speed track memory bandwidth almost exactly — a card with twice the bandwidth generates roughly twice as fast.

Its internal architecture is not on file, so memory is approximated from the parameter count and marked accordingly. Expect the real figure to differ, more so at long context.

What went into building it

Producing it required arithmetic totalling around 2.1 × 10²² FLOP, on hardware recorded as NVIDIA Tesla V100 DGXS 32 GB. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 167,000,000,000 tokens of text.

The reason it appears in this catalogue at all: highly cited,SOTA improvement.

Step by step

How to choose a GPU for XLM-RoBERTa

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

01

Check what it needs before anything else

Start from what it actually needs, which is the requirement of XLM-RoBERTa, needing around 1.3 GB at a compression of Q8_0. Capacity is the gate — a card either holds it or it does not.
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 XLM-RoBERTa.
03

Set a quality floor

Compression is what makes a model fit smaller cards, at some cost in accuracy, reaching a compression of Q8_0 on the smallest card that fits. 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.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for XLM-RoBERTa. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 6,160 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage it from those with room to spare, in the case of XLM-RoBERTa. 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.
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 you have settled on XLM-RoBERTa.

Answers

XLM-RoBERTa — common questions

XLM-RoBERTa— who created it?

It was published by Facebook AI, based in United States of America, an organisation categorised as industry.

XLM-RoBERTa— when was it released?

It was published in November 2019. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

XLM-RoBERTa— what is it used for?

It works in the domain of Language, and is recorded as handling the task of named entity recognition (NER), Question answering, Text classification. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

XLM-RoBERTa— 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.

XLM-RoBERTa— how much compute was used to train it?

Training consumed around 2.1 × 10²² FLOP, on hardware recorded as 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.

XLM-RoBERTa— 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. Every figure here assumes the whole model is resident on the card.

XLM-RoBERTa— 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: 818. So a second card is rarely the answer here.

XLM-RoBERTa— why does the quantisation differ between cards?

Each card is shown running the least-compressed copy it can hold. The number of distinct compression levels across the cards that fit it: 1. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

XLM-RoBERTa— how accurate are these speed estimates?

These are estimates with real error bars, and any of them could reasonably land anywhere in its published range depending on which runtime you use. The fastest result here: 3,696–9,857 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

XLM-RoBERTa— what GPU do I need to run it?

The smallest card in our catalogue that holds it is Tesla C1080, with a memory capacity of 4 GB. It runs the model at a compression of Q8_0 using about 1.3 GB, and produces roughly 67.0 tokens per second. The number of cards able to run it in total: 818.

XLM-RoBERTa— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 6,160 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and the number of cards clearing that: 809.

XLM-RoBERTa— how much VRAM does it need?

It needs about 1.3 GB at a compression of Q8_0, 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.

XLM-RoBERTa— can I run it on a GPU holding 8 GB?

Yes. The card CMP 170HX 8 GB, holding 8 GB, runs it at a compression of Q8_0, using about 1.3 GB and generating roughly 1,147 tokens per second. The fit is comfortable.

XLM-RoBERTa— can I run it on a GPU holding 12 GB?

Yes. The card GeForce RTX 3080 Ti, holding 12 GB, runs it at a compression of Q8_0, using about 1.3 GB and generating roughly 703 tokens per second. The fit is comfortable.

XLM-RoBERTa— can I run it on a GPU holding 16 GB?

Yes. The card Tesla V100 SXM2 16 GB, holding 16 GB, runs it at a compression of Q8_0, using about 1.3 GB and generating roughly 870 tokens per second. The fit is comfortable.

XLM-RoBERTa— can I run it on a GPU holding 24 GB?

Yes. The card GeForce RTX 5090 D V2, holding 24 GB, runs it at a compression of Q8_0, using about 1.3 GB and generating roughly 1,032 tokens per second. The fit is comfortable.

XLM-RoBERTa— 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.

XLM-RoBERTa— how many parameters does it have?

It has a parameter count of 550M. The number of parameters in the model is specified as "550M params" for XLM-R. 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.

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