ESM2-650M TPS calculator

Open weights Meta AI,New York University (NYU),Stanford University,Massachusetts Institute of Technology (MIT) 650M 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

818 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Tesla C1080

4 GB · Q8_0 · 56.7 tok/s

Fastest card

B200

5,213 tok/s · 180 GB

Which GPUs can run ESM2-650M?

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
5,213 tok/s 3,128–8,340 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.4 GB	Q8_0	Comfortable
5,213 tok/s 3,128–8,340 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.4 GB	Q8_0	Comfortable
4,162 tok/s 2,497–6,660 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.4 GB	Q8_0	Comfortable
4,162 tok/s 2,497–6,660 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.4 GB	Q8_0	Comfortable
3,329 tok/s 1,997–5,326 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.4 GB	Q8_0	Comfortable
3,186 tok/s 1,912–5,098 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.4 GB	Q8_0	Comfortable
3,186 tok/s 1,912–5,098 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.4 GB	Q8_0	Comfortable
3,049 tok/s 1,830–4,879 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.4 GB	Q8_0	Comfortable
2,706 tok/s 1,624–4,330 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.4 GB	Q8_0	Comfortable
2,706 tok/s 1,624–4,330 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.4 GB	Q8_0	Comfortable
2,706 tok/s 1,624–4,330 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.4 GB	Q8_0	Comfortable
2,567 tok/s 1,540–4,108 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.4 GB	Q8_0	Comfortable
2,189 tok/s 1,314–3,503 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.4 GB	Q8_0	Comfortable
2,189 tok/s 1,314–3,503 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.4 GB	Q8_0	Comfortable
2,189 tok/s 1,314–3,503 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.4 GB	Q8_0	Comfortable
2,189 tok/s 1,314–3,503 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.4 GB	Q8_0	Comfortable
2,189 tok/s 1,314–3,503 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.4 GB	Q8_0	Comfortable
1,667 tok/s 1,000–2,667 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.4 GB	Q8_0	Comfortable
1,667 tok/s 1,000–2,667 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.4 GB	Q8_0	Comfortable
1,389 tok/s 834–2,223 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.4 GB	Q8_0	Comfortable
1,360 tok/s 816–2,175 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.4 GB	Q8_0	Comfortable
1,329 tok/s 798–2,127 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.4 GB	Q8_0	Comfortable
1,329 tok/s 798–2,127 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.4 GB	Q8_0	Comfortable
1,329 tok/s 798–2,127 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.4 GB	Q8_0	Comfortable
1,329 tok/s 798–2,127 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	1.4 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: Meta AI,New York University (NYU),Stanford University,Massachusetts Institute of Technology (MIT)
Organisation type: Industry,Academia,Academia,Academia
Country: United States of America
Published: 21 July 2022
Authors: Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Nikita Smetanin, Robert Verkuil, Ori Kabeli, Yaniv Shmueli, Allan dos Santos Costa, Maryam Fazel-Zarandi, 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: Proteins, Protein or nucleotide language model (pLM/nLM), Protein folding prediction
Approach: Unsupervised

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: 650M
Training data: 15,360,000,000 tokens
Epochs: 83.3

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: 7.6 × 10²¹ FLOP
How it was established: Hardware,Third-party estimation

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: 512
Wall-clock time: 192 hours (8 days)
Power draw: 308.0 kW
Compute cost: $20,972

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: 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
Citations: 636

Sources

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

Reference: Evolutionary-scale prediction of atomic-level protein structure with a language model
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run ESM2-650M

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 ESM2-650M

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

1.4 GB

Fastest

5,213 tok/s

ESM2-650M is small enough at 650M 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 56.7 tokens per second.

A B200 is the fastest we calculate for it: about 5,213 tokens per second, from 8,000 GB/s of memory bandwidth.

What this model is

ESM2-650M was published by Meta AI,New York University (NYU),Stanford University,Massachusetts Institute of Technology (MIT), in United States of America, in July 2022. industry,Academia,Academia,Academia is the category the publisher falls under.

It works in Biology, and is recorded as doing proteins, Protein or nucleotide language model (pLM/nLM), Protein folding prediction.

Published weights mean the model runs on your machine rather than someone else's, which is what makes the hardware question below answerable at all.

What decides the speed

Across every card that can run it, the middle of the range is about 146.4 tokens per second, and 809 of them clear the ten tokens per second that roughly matches reading speed.

Being dense, it reads all of itself per token, which is why the ordering by speed below follows the ordering by memory bandwidth so closely.

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 7.6 × 10²¹ FLOP of arithmetic, on NVIDIA V100, which is a statement about the training budget rather than about inference.

It was trained on about 15,360,000,000 tokens of text.

Step by step

How to choose a GPU for ESM2-650M

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

The table lists every card that can hold ESM2-650M — around 1.4 GB at Q8_0. That figure, not the card's headline performance, is what decides whether it runs.
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 ESM2-650M.
03

Choose how far you will compress it

The quantisation column varies by card, because a bigger card holds a more accurate copy of ESM2-650M — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Sort by speed

Ranking by tokens per second for ESM2-650M follows memory bandwidth, not core counts, which is why the B200 tops it at 5,213 tok/s.
05

Check the fit verdict before buying

The fit column separates cards that just manage ESM2-650M from those with room to spare. Buy for the second if the context might grow.
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 ESM2-650M.

Answers

ESM2-650M — common questions

When was ESM2-650M released?

ESM2-650M 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.

What is ESM2-650M used for?

ESM2-650M works in Biology, and is recorded as handling proteins, Protein or nucleotide language model (pLM/nLM), Protein folding prediction. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download ESM2-650M?

The weights for ESM2-650M 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 ESM2-650M?

Around 7.6 × 10²¹ 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 ESM2-650M 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 ESM2-650M is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run ESM2-650M faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold ESM2-650M on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for ESM2-650M?

Each card is shown running the least-compressed copy it can hold, and ESM2-650M appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

How accurate are these ESM2-650M speed estimates?

These are estimates with real error bars. The fastest result here, 3,128–8,340 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 ESM2-650M?

The smallest card in our catalogue that holds ESM2-650M is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.4 GB, and produces roughly 56.7 tokens per second. 818 cards in total can run it.

How fast is ESM2-650M on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 5,213 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 809 of the cards that can run ESM2-650M clear that.

How much VRAM does ESM2-650M need?

About 1.4 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 ESM2-650M on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 1.4 GB and generating roughly 971 tokens per second — a comfortable fit.

Can I run ESM2-650M on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 1.4 GB and generating roughly 595 tokens per second — a comfortable fit.

Can I run ESM2-650M on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 1.4 GB and generating roughly 736 tokens per second — a comfortable fit.

Can I run ESM2-650M on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 1.4 GB and generating roughly 873 tokens per second — a comfortable fit.

Is ESM2-650M open source?

Its weights are published, so ESM2-650M 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 ESM2-650M have?

ESM2-650M has 650M parameters. In the name. 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 ESM2-650M?

ESM2-650M was published by Meta AI,New York University (NYU),Stanford University,Massachusetts Institute of Technology (MIT), based in United States of America, categorised as industry,Academia,Academia,Academia.

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