NLLB TPS calculator

Open weights Meta AI 54.5B parameters July 2022

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

92 cards that can run it

818 cards we hold specifications for

Smallest card that fits

FirePro S9170

32 GB · Q3_K_M · 5.2 tok/s

Fastest card

B200

62.2 tok/s · 180 GB

Which GPUs can run NLLB?

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

92 cards match

Calculating

					Needs	Quantisation	Fit
62.2 tok/s 37–99 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	59.0 GB	Q8_0	Comfortable
62.2 tok/s 37–99 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	59.0 GB	Q8_0	Comfortable
49.6 tok/s 30–79 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	59.0 GB	Q8_0	Comfortable
49.6 tok/s 30–79 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	59.0 GB	Q8_0	Comfortable
39.7 tok/s 24–64 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	59.0 GB	Q8_0	Comfortable
39.2 tok/s 24–63 · low confidence	DRIVE A100 PROD NVIDIA	32 GB	1,870 GB/s	May 2020	27.3 GB	Q3_K_M	Tight
39.2 tok/s 24–63 · low confidence	GRID A100A NVIDIA	32 GB	1,870 GB/s	May 2020	27.3 GB	Q3_K_M	Tight
38.0 tok/s 23–61 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	59.0 GB	Q8_0	Comfortable
38.0 tok/s 23–61 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	59.0 GB	Q8_0	Comfortable
37.5 tok/s 23–60 · low confidence	GeForce RTX 5090 NVIDIA	32 GB	1,790 GB/s	Jan 2025	27.3 GB	Q3_K_M	Tight
37.5 tok/s 23–60 · low confidence	GeForce RTX 5090 D NVIDIA	32 GB	1,790 GB/s	Jan 2025	27.3 GB	Q3_K_M	Tight
36.4 tok/s 22–58 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	59.0 GB	Q8_0	Comfortable
32.3 tok/s 19–52 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	59.0 GB	Q8_0	Comfortable
32.3 tok/s 19–52 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	59.0 GB	Q8_0	Comfortable
32.3 tok/s 19–52 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	59.0 GB	Q8_0	Comfortable
30.6 tok/s 18–49 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	59.0 GB	Q8_0	Comfortable
28.0 tok/s 17–45 · low confidence	A100 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Jun 2020	33.7 GB	Q4_K_M	Tight
28.0 tok/s 17–45 · low confidence	A100 SXM4 40 GB NVIDIA	40 GB	1,560 GB/s	May 2020	33.7 GB	Q4_K_M	Tight
28.0 tok/s 17–45 · low confidence	A800 PCIe 40 GB NVIDIA	40 GB	1,560 GB/s	Nov 2022	33.7 GB	Q4_K_M	Tight
26.1 tok/s 16–42 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	59.0 GB	Q8_0	Comfortable
26.1 tok/s 16–42 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	59.0 GB	Q8_0	Comfortable
26.1 tok/s 16–42 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	59.0 GB	Q8_0	Comfortable
26.1 tok/s 16–42 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	59.0 GB	Q8_0	Comfortable
26.1 tok/s 16–42 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	59.0 GB	Q8_0	Comfortable
26.0 tok/s 16–42 · low confidence	GRID A100B NVIDIA	48 GB	1,870 GB/s	May 2020	40.0 GB	Q5_K_M	Tight

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: Meta AI
Organisation type: Industry
Country: United States of America
Published: 6 July 2022
Authors: Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Prangthip Hansanti, John Hoffman, Semarley Jarrett, Kaushik Ram Sadagopan, Dirk Rowe, Shannon Spruit, Chau Tran, Pierre Andrews, Necip Fazi…

What it does

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

Domain: Language
Task: Translation
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: 54.5B
Training data: 300,000,000,000 tokens
Batch size: 1,000,000

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: 1.8 × 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 A100 SXM4 80 GB
Chip-hours: 59,168
Compute cost: $50,667

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: SOTA improvement
Record confidence: Confident
Citations: 1,569

Sources

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

Reference: No Language Left Behind: Scaling Human-Centered Machine Translation
Last updated: 1 January 2026

The extremes

The ten fastest GPUs that run NLLB

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 NLLB

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

Hardware requirements in practice

Minimum card

FirePro S9170

Memory needed

27.3 GB

Fastest

62.2 tok/s

NLLB reaches a parameter count of 54.5B. That lands in the range a serious desktop card can handle once the weights are compressed. The number of cards we track that can run it: 92.

The entry point is FirePro S9170, with a memory capacity of 32 GB, running it at a compression of Q3_K_M and producing around 5.2 tokens per second.

The quickest result comes from B200, generating roughly 62.2 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

About this model

NLLB was published by Meta AI, in the country recorded as United States of America, during July 2022. The category the publisher falls under is industry.

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

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

How fast it runs, and why

The median result is around 15.9 tokens per second. Exceeding reading speed outright: 72 of them.

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

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.

How it was trained

The training run consumed about 1.8 × 10²² FLOP, on hardware recorded as NVIDIA A100 SXM4 80 GB. That figure measures what producing the model cost, and has no bearing on how fast it answers.

It was trained on a corpus of about 300,000,000,000 tokens of text.

Its inclusion criterion: sOTA improvement.

Step by step

How to choose a GPU for NLLB

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

Every card here has been checked against NLLB, needing around 27.3 GB at a compression of Q3_K_M. Capacity is the gate — a card either holds it or it does not.
02

Match the context to your actual use

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 NLLB.
03

Choose how far you will compress it

Compression is what makes a model fit smaller cards, at some cost in accuracy, reaching a compression of Q3_K_M 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

Rank by throughput rather than spec sheet

Ranking by tokens per second follows memory bandwidth rather than core counts, for NLLB. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 62.2 tok/s.
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 NLLB. 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

Check the card from the other side

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 NLLB.

Answers

NLLB — common questions

NLLB— how many parameters does it have?

It has a parameter count of 54.5B. Section 8.2.4: "The model has a total of 54.5B parameters and FLOPs similar to that of a 3.3B dense model". 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.

NLLB— who created it?

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

NLLB— when was it released?

It was published in July 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

NLLB— what is it used for?

It works in the domain of Language, and is recorded as handling the task of translation. 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.

NLLB— 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.

NLLB— how much compute was used to train it?

Training consumed around 1.8 × 10²² FLOP, on hardware recorded as NVIDIA A100 SXM4 80 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.

NLLB— can I run it if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part held in system memory drags the whole thing down. The nearest miss we calculate falls short by 8.5 GB. Every figure here assumes the whole model is resident on the card.

NLLB— 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: 92. So a second card is rarely the answer here.

NLLB— why does the quantisation differ between cards?

Because capacity varies, so does how hard it has to be squeezed. The number of distinct levels in the table above: 5. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

NLLB— 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: 37–99 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

NLLB— what GPU do I need to run it?

The smallest card in our catalogue that holds it is FirePro S9170, with a memory capacity of 32 GB. It runs the model at a compression of Q3_K_M using about 27.3 GB, and produces roughly 5.2 tokens per second. The number of cards able to run it in total: 92.

NLLB— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 62.2 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: 72.

NLLB— how much VRAM does it need?

It needs about 27.3 GB at a compression of Q3_K_M, 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.

NLLB— 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.

Source

Original publication

Record last updated 1 January 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