ESM3-open-small TPS calculator

Open weights EvolutionaryScale,University of California (UC) Berkeley 1.4B parameters June 2024

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 · 26.3 tok/s

Fastest card

B200

2,420 tok/s · 180 GB

Which GPUs can run ESM3-open-small?

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
2,420 tok/s 1,452–3,872 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	2.2 GB	Q8_0	Comfortable
2,420 tok/s 1,452–3,872 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	2.2 GB	Q8_0	Comfortable
1,933 tok/s 1,160–3,092 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	2.2 GB	Q8_0	Comfortable
1,933 tok/s 1,160–3,092 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	2.2 GB	Q8_0	Comfortable
1,546 tok/s 927–2,473 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	2.2 GB	Q8_0	Comfortable
1,479 tok/s 888–2,367 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	2.2 GB	Q8_0	Comfortable
1,479 tok/s 888–2,367 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	2.2 GB	Q8_0	Comfortable
1,416 tok/s 849–2,265 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	2.2 GB	Q8_0	Comfortable
1,257 tok/s 754–2,010 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	2.2 GB	Q8_0	Comfortable
1,257 tok/s 754–2,010 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	2.2 GB	Q8_0	Comfortable
1,257 tok/s 754–2,010 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	2.2 GB	Q8_0	Comfortable
1,192 tok/s 715–1,907 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	2.2 GB	Q8_0	Comfortable
1,016 tok/s 610–1,626 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	2.2 GB	Q8_0	Comfortable
1,016 tok/s 610–1,626 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	2.2 GB	Q8_0	Comfortable
1,016 tok/s 610–1,626 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	2.2 GB	Q8_0	Comfortable
1,016 tok/s 610–1,626 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	2.2 GB	Q8_0	Comfortable
1,016 tok/s 610–1,626 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	2.2 GB	Q8_0	Comfortable
774 tok/s 464–1,238 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	2.2 GB	Q8_0	Comfortable
774 tok/s 464–1,238 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	2.2 GB	Q8_0	Comfortable
645 tok/s 387–1,032 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	2.2 GB	Q8_0	Comfortable
631 tok/s 379–1,010 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	2.2 GB	Q8_0	Comfortable
617 tok/s 370–987 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	2.2 GB	Q8_0	Comfortable
617 tok/s 370–987 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	2.2 GB	Q8_0	Comfortable
617 tok/s 370–987 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	2.2 GB	Q8_0	Comfortable
617 tok/s 370–987 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	2.2 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: EvolutionaryScale,University of California (UC) Berkeley
Organisation type: Industry,Academia
Country: United States of America
Published: 25 June 2024
Authors: Thomas Hayes, Roshan Rao, Halil Akin, Nicholas James Sofroniew, Deniz Oktay, Zeming Lin, Robert Verkuil, Vincent Quy Tran, Jonathan Deaton, Marius Wiggert, Rohil Badkundri, Irhum Shafkat, Jun Gong, Alexander Derry, Raul Santiago Molina, Neil Thomas, Yousuf Khan, Chetan Mishra, Carolyn Kim, Liam J Bartie, Patrick D Hsu, Tom Sercu, Salvatore Candido, Alexander Rives

What it does

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

Domain: Biology
Task: Protein generation

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: 1.4B
Training data: 48,000,000,000 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: 2.7 × 10²¹ FLOP
How it was established: Reported

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: Unreleased

How it is classified

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

Record confidence: Confident

Sources

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

Reference: ESM3: Simulating 500 million years of evolution with a language model
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run ESM3-open-small

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 ESM3-open-small

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

2.2 GB

Fastest

2,420 tok/s

ESM3-open-small is small enough at 1.4B parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The least hardware that works is a Tesla C1080. Its 4 GB is enough at Q8_0 compression, giving roughly 26.3 tokens per second.

At the other end, a B200 generates roughly 2,420 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

About this model

ESM3-open-small was published by EvolutionaryScale,University of California (UC) Berkeley, in United States of America, in June 2024. industry,Academia is the category the publisher falls under.

It works in Biology, and is recorded as doing protein generation.

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.

How fast it runs, and why

The median result is around 68.0 tokens per second; 797 cards produce text faster than most people read it.

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

Training it took roughly 2.7 × 10²¹ FLOP of computation — a measure of what producing the model cost, not of how fast it answers.

Around 48,000,000,000 tokens went into training it.

Step by step

How to choose a GPU for ESM3-open-small

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

Look at what ESM3-open-small actually needs — around 2.2 GB at Q8_0. No amount of processing power compensates for a card that cannot hold it.
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: at long context ESM3-open-small can slip off a card that handles short questions easily.
03

Choose how far you will compress it

The quantisation column varies by card, because a bigger card holds a more accurate copy of ESM3-open-small — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Sort by speed

The speed ordering for ESM3-open-small is effectively an ordering by memory bandwidth, which is why the B200 tops it at 2,420 tok/s.
05

Look at the headroom, not just the fit

A tight fit runs ESM3-open-small but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
06

Check the card from the other side

Each card page repeats this sweep for every model we hold. It answers what else the hardware is good for, beyond ESM3-open-small.

Answers

ESM3-open-small — common questions

How many parameters does ESM3-open-small have?

ESM3-open-small has 1.4B parameters. 1.4 billion. 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 ESM3-open-small?

ESM3-open-small was published by EvolutionaryScale,University of California (UC) Berkeley, based in United States of America, categorised as industry,Academia.

When was ESM3-open-small released?

ESM3-open-small was published in June 2024. 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 ESM3-open-small used for?

ESM3-open-small works in Biology, and is recorded as handling protein generation. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download ESM3-open-small?

The weights for ESM3-open-small 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 ESM3-open-small?

Around 2.7 × 10²¹ FLOP. 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 ESM3-open-small 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 ESM3-open-small is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run ESM3-open-small faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run ESM3-open-small alone, the case for pairing is weak.

Why does the quantisation differ between cards for ESM3-open-small?

A larger card holds a more accurate copy. Across the cards that run ESM3-open-small, 1 compression levels are used; the floor control above pins it to one.

How accurate are these ESM3-open-small speed estimates?

These are estimates with real error bars. The fastest result here, 1,452–3,872 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 ESM3-open-small?

The smallest card in our catalogue that holds ESM3-open-small is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 2.2 GB, and produces roughly 26.3 tokens per second. 818 cards in total can run it.

How fast is ESM3-open-small on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 2,420 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 797 of the cards that can run ESM3-open-small clear that.

How much VRAM does ESM3-open-small need?

About 2.2 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 ESM3-open-small on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 2.2 GB and generating roughly 451 tokens per second — a comfortable fit.

Can I run ESM3-open-small on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 2.2 GB and generating roughly 276 tokens per second — a comfortable fit.

Can I run ESM3-open-small on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 2.2 GB and generating roughly 342 tokens per second — a comfortable fit.

Can I run ESM3-open-small on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 2.2 GB and generating roughly 405 tokens per second — a comfortable fit.

Is ESM3-open-small open source?

Its weights are published, so ESM3-open-small 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 28 November 2025

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