ProteinBERT TPS calculator

Open weights Hebrew University of Jerusalem,Ben-Gurion University of the Negev,Deep Trading 16M parameters February 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

818 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Tesla C1080

4 GB · Q8_0 · 2,304 tok/s

Fastest card

B200

211,765 tok/s · 180 GB

Which GPUs can run ProteinBERT?

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
211,765 tok/s 127,059–338,824 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	0.7 GB	Q8_0	Comfortable
211,765 tok/s 127,059–338,824 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	0.7 GB	Q8_0	Comfortable
169,099 tok/s 101,460–270,559 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
169,099 tok/s 101,460–270,559 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
135,238 tok/s 81,143–216,381 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
129,441 tok/s 77,665–207,106 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
129,441 tok/s 77,665–207,106 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
123,882 tok/s 74,329–198,212 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	0.7 GB	Q8_0	Comfortable
109,946 tok/s 65,967–175,913 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
109,946 tok/s 65,967–175,913 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
109,946 tok/s 65,967–175,913 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
104,294 tok/s 62,576–166,871 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
88,941 tok/s 53,365–142,306 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
88,941 tok/s 53,365–142,306 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	0.7 GB	Q8_0	Comfortable
88,941 tok/s 53,365–142,306 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
88,941 tok/s 53,365–142,306 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
88,941 tok/s 53,365–142,306 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
67,722 tok/s 40,633–108,356 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
67,722 tok/s 40,633–108,356 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
56,435 tok/s 33,861–90,296 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
55,231 tok/s 33,139–88,369 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
54,000 tok/s 32,400–86,400 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	0.7 GB	Q8_0	Comfortable
54,000 tok/s 32,400–86,400 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	0.7 GB	Q8_0	Comfortable
54,000 tok/s 32,400–86,400 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	0.7 GB	Q8_0	Comfortable
54,000 tok/s 32,400–86,400 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	0.7 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: Hebrew University of Jerusalem,Ben-Gurion University of the Negev,Deep Trading
Organisation type: Academia,Academia,Industry
Country: Israel, United States of America
Published: 10 February 2022
Authors: Nadav Brandes, Dan Ofer, Yam Peleg, Nadav Rappoport, Michal Linial

What it does

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

Domain: Biology
Task: Proteins, Protein generation, Protein representation learning
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: 16M
Training data: 37,598,200,000 tokens
Epochs: 6.4
Batch size: 26,008

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: 6.5 × 10¹⁹ 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 Quadro RTX 5000
Chips used: 1
Wall-clock time: 672 hours (28 days)
Power draw: 254 W

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: 775

Sources

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

Reference: ProteinBERT: a universal deep-learning model of protein sequence and function
Last updated: 1 January 2026

The extremes

The ten fastest GPUs that run ProteinBERT

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 ProteinBERT

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

0.7 GB

Fastest

211,765 tok/s

ProteinBERT is small enough at 16M 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 2,304 tokens per second.

At the other end, a B200 generates roughly 211,765 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

What this model is

ProteinBERT was published by Hebrew University of Jerusalem,Ben-Gurion University of the Negev,Deep Trading, in Israel, in February 2022. The organisation is categorised as academia,Academia,Industry.

It works in Biology, and is recorded as doing proteins, Protein generation, Protein representation learning.

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 5,946.4 tokens per second, and 818 exceed reading speed outright.

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.

Training and provenance

Producing it required around 6.5 × 10¹⁹ FLOP of arithmetic, on NVIDIA Quadro RTX 5000, which is a statement about the training budget rather than about inference.

The training set ran to roughly 37,598,200,000 tokens.

Its inclusion criterion is sOTA improvement.

Step by step

How to choose a GPU for ProteinBERT

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 ProteinBERT — around 0.7 GB at Q8_0. Capacity is the gate — a card either holds it or it does not.
02

Decide how long your conversations run

Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason ProteinBERT stops fitting a card that seemed fine.
03

Decide how much compression you will accept

The quantisation column varies by card, because a bigger card holds a more accurate copy of ProteinBERT — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Rank by throughput rather than spec sheet

The speed ordering for ProteinBERT is effectively an ordering by memory bandwidth, which is why the B200 tops it at 211,765 tok/s.
05

Read the fit column last

Tight means ProteinBERT loads and works, with no room to raise the context later. Comfortable means you can. The difference matters more than a few tokens per second.
06

Open the card you have settled on

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for ProteinBERT alone — a card is usually bought for more than one model.

Answers

ProteinBERT — common questions

Why does the quantisation differ between cards for ProteinBERT?

Each card is shown running the least-compressed copy it can hold, and ProteinBERT appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

How accurate are these ProteinBERT speed estimates?

These are estimates with real error bars. The fastest result here, 127,059–338,824 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

What GPU do I need to run ProteinBERT?

The smallest card in our catalogue that holds ProteinBERT is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 0.7 GB, and produces roughly 2,304 tokens per second. 818 cards in total can run it.

How fast is ProteinBERT on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 211,765 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 818 of the cards that can run ProteinBERT clear that.

How much VRAM does ProteinBERT need?

About 0.7 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 ProteinBERT on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 0.7 GB and generating roughly 39,441 tokens per second — a comfortable fit.

Can I run ProteinBERT on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 0.7 GB and generating roughly 24,152 tokens per second — a comfortable fit.

Can I run ProteinBERT on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 0.7 GB and generating roughly 29,912 tokens per second — a comfortable fit.

Can I run ProteinBERT on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 0.7 GB and generating roughly 35,471 tokens per second — a comfortable fit.

Is ProteinBERT open source?

Its weights are published, so ProteinBERT 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.

How many parameters does ProteinBERT have?

ProteinBERT has 16M parameters. "Altogether, it includes ∼16M trainable parameters, making it substantially smaller than other protein language models". 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 ProteinBERT?

ProteinBERT was published by Hebrew University of Jerusalem,Ben-Gurion University of the Negev,Deep Trading, based in Israel, categorised as academia,Academia,Industry.

When was ProteinBERT released?

ProteinBERT was published in February 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.

What is ProteinBERT used for?

ProteinBERT works in Biology, and is recorded as handling proteins, Protein generation, Protein representation learning. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download ProteinBERT?

The weights for ProteinBERT 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 ProteinBERT?

Around 6.5 × 10¹⁹ FLOP, on NVIDIA Quadro RTX 5000. 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 ProteinBERT if it does not fit in my GPU?

It can be split between the card and system memory, but ProteinBERT generates painfully slowly that way. Nothing on this page assumes offloading.

Would two GPUs run ProteinBERT faster?

Two cards buy memory rather than speed. That matters for ProteinBERT only if one card cannot hold it — 818 can, so a second adds little.

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