ESM1-670M (UR50/D) TPS calculator

Open weights Facebook AI Research,New York University (NYU) 669.2M parameters August 2020

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 · 55.1 tok/s

Fastest card

B200

5,063 tok/s · 180 GB

Which GPUs can run ESM1-670M (UR50/D)?

Set the inputs, read the answer

A longer conversation needs more memory, which can push this model off smaller cards.

Hides cards that would only fit the model by compressing it below this point.

818 cards match

Calculating
Needs Quantisation Fit
5,063 tok/s

3,038–8,101 · low confidence

B200 NVIDIA 180 GB 8,000 GB/s Jan 2024 1.4 GB Q8_0 Comfortable
5,063 tok/s

3,038–8,101 · low confidence

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 1.4 GB Q8_0 Comfortable
4,043 tok/s

2,426–6,469 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 1.4 GB Q8_0 Comfortable
4,043 tok/s

2,426–6,469 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 1.4 GB Q8_0 Comfortable
3,233 tok/s

1,940–5,173 · low confidence

Radeon Instinct MI300 AMD 128 GB 6,550 GB/s Jan 2023 1.4 GB Q8_0 Comfortable
3,095 tok/s

1,857–4,952 · low confidence

H200 NVL NVIDIA 141 GB 4,890 GB/s Nov 2024 1.4 GB Q8_0 Comfortable
3,095 tok/s

1,857–4,952 · low confidence

H200 SXM 141 GB NVIDIA 141 GB 4,890 GB/s Nov 2024 1.4 GB Q8_0 Comfortable
2,962 tok/s

1,777–4,739 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 1.4 GB Q8_0 Comfortable
2,629 tok/s

1,577–4,206 · low confidence

Radeon Instinct MI300A AMD 128 GB 5,325 GB/s Dec 2023 1.4 GB Q8_0 Comfortable
2,629 tok/s

1,577–4,206 · low confidence

Radeon Instinct MI300X AMD 192 GB 5,325 GB/s Dec 2023 1.4 GB Q8_0 Comfortable
2,629 tok/s

1,577–4,206 · low confidence

Radeon Instinct MI308X AMD 192 GB 5,325 GB/s Dec 2023 1.4 GB Q8_0 Comfortable
2,494 tok/s

1,496–3,990 · low confidence

H100 NVL 94 GB NVIDIA 94 GB 3,940 GB/s Mar 2023 1.4 GB Q8_0 Comfortable
2,127 tok/s

1,276–3,402 · low confidence

H100 PCIe 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 1.4 GB Q8_0 Comfortable
2,127 tok/s

1,276–3,402 · low confidence

H100 SXM5 80 GB NVIDIA 80 GB 3,360 GB/s Oct 2022 1.4 GB Q8_0 Comfortable
2,127 tok/s

1,276–3,402 · low confidence

H100 SXM5 94 GB NVIDIA 94 GB 3,360 GB/s Mar 2023 1.4 GB Q8_0 Comfortable
2,127 tok/s

1,276–3,402 · low confidence

H100 SXM5 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 1.4 GB Q8_0 Comfortable
2,127 tok/s

1,276–3,402 · low confidence

H800 SXM5 NVIDIA 80 GB 3,360 GB/s Mar 2023 1.4 GB Q8_0 Comfortable
1,619 tok/s

972–2,591 · low confidence

Radeon Instinct MI250 AMD 128 GB 3,280 GB/s Nov 2021 1.4 GB Q8_0 Comfortable
1,619 tok/s

972–2,591 · low confidence

Radeon Instinct MI250X AMD 128 GB 3,280 GB/s Nov 2021 1.4 GB Q8_0 Comfortable
1,349 tok/s

810–2,159 · low confidence

Data Center GPU Max 1550 Intel 128 GB 3,280 GB/s Jan 2023 1.4 GB Q8_0 Comfortable
1,321 tok/s

792–2,113 · low confidence

Data Center GPU Max Subsystem Intel 128 GB 3,210 GB/s Jan 2023 1.4 GB Q8_0 Comfortable
1,291 tok/s

775–2,066 · low confidence

A100 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Nov 2020 1.4 GB Q8_0 Comfortable
1,291 tok/s

775–2,066 · low confidence

A100X NVIDIA 80 GB 2,040 GB/s Jun 2021 1.4 GB Q8_0 Comfortable
1,291 tok/s

775–2,066 · low confidence

A800 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Aug 2022 1.4 GB Q8_0 Comfortable
1,291 tok/s

775–2,066 · 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
Facebook AI Research,New York University (NYU)
Organisation type
Industry,Academia
Country
United States of America, France
Published
31 August 2020
Authors
Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus

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)

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
669.2M

See Table 1

Training data
33,847,900,000 tokens

"(iii) UR50/D, 124.9M UniRef50 cluster members sampled evenly by cluster" As a rule of thumb, 200 amino acids per protein, so 124.9M * 200 = 24.98 billion Epochs: 906k steps * 131,072 tokens/batch / 24.98B = 4.75 epochs

Epochs
3.51
Batch size
131,072

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
4.8 × 10²⁰ FLOP

Information: 128 NVIDIA V100 GPUs [Pre-training details] 906k steps [See Table S2: Hyperparameters] 131,072 tokens per batch ["We trained with 131,072 tokens per batch (128 gpus x 1024 tokens)." - Pre-training details] Estimate: 906e3 updates * 3 * 131072 tokens/update * 2 * 669.2e6 parameters = 4.8e20 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 V100
Chips used
128
Power draw
78.2 kW
Compute cost
$1,055

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

MIT. looks like inference code, not training code: https://github.com/facebookresearch/esm

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
Highly cited

"We apply the representations to a range of prediction tasks and find that they improve state-of-art features across the applications."

Record confidence
Confident
Citations
2,637

Sources

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

Reference
Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences
Last updated
11 February 2026

The extremes

The ten fastest GPUs that run ESM1-670M (UR50/D)

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.

  1. 01 B300 288 GB · 8,000 GB/s · Q8_0 5,063 tok/s
  2. 02 B200 180 GB · 8,000 GB/s · Q8_0 5,063 tok/s
  3. 03 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q8_0 4,043 tok/s
  4. 04 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q8_0 4,043 tok/s
  5. 05 Radeon Instinct MI300 128 GB · 6,550 GB/s · Q8_0 3,233 tok/s
  6. 06 H200 NVL 141 GB · 4,890 GB/s · Q8_0 3,095 tok/s
  7. 07 H200 SXM 141 GB 141 GB · 4,890 GB/s · Q8_0 3,095 tok/s
  8. 08 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q8_0 2,962 tok/s
  9. 09 Radeon Instinct MI300A 128 GB · 5,325 GB/s · Q8_0 2,629 tok/s
  10. 10 Radeon Instinct MI300X 192 GB · 5,325 GB/s · Q8_0 2,629 tok/s

The smallest GPUs that still run ESM1-670M (UR50/D)

The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.

  1. 01 GeForce RTX 4010 4 GB · needs 1.4 GB · Q8_0 · comfortable 60.8 tok/s
  2. 02 RTX A400 4 GB · needs 1.4 GB · Q8_0 · comfortable 60.8 tok/s
  3. 03 RTX 500 Mobile Ada Generation 4 GB · needs 1.4 GB · Q8_0 · comfortable 81.0 tok/s
  4. 04 GeForce RTX 3050 A Mobile 4 GB · needs 1.4 GB · Q8_0 · comfortable 122 tok/s
  5. 05 Jetson Orin Nano 4 GB 4 GB · needs 1.4 GB · Q8_0 · comfortable 21.6 tok/s
  6. 06 Radeon RX 6450M 4 GB · needs 1.4 GB · Q8_0 · comfortable 63.2 tok/s
  7. 07 Radeon RX 6550M 4 GB · needs 1.4 GB · Q8_0 · comfortable 71.1 tok/s
  8. 08 Radeon RX 6550S 4 GB · needs 1.4 GB · Q8_0 · comfortable 63.2 tok/s
  9. 09 Arc A310 4 GB · needs 1.4 GB · Q8_0 · comfortable 51.0 tok/s
  10. 10 Arc Pro A30M 4 GB · needs 1.4 GB · Q8_0 · comfortable 52.7 tok/s

What the numbers mean

What it takes to run this model

Minimum card

Tesla C1080

Memory needed

1.4 GB

Fastest

5,063 tok/s

ESM1-670M (UR50/D) is small enough at 669.2M parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The smallest card that holds it is the Tesla C1080 with 4 GB, running it at Q8_0 and producing around 55.1 tokens per second.

Top of the range is the B200, at roughly 5,063 tokens per second thanks to 8,000 GB/s of bandwidth.

Background

ESM1-670M (UR50/D) was published by Facebook AI Research,New York University (NYU), in United States of America, in August 2020. industry,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).

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

The median result is around 142.2 tokens per second; 809 cards produce text faster than most people read it.

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.

Memory here is estimated from size rather than computed from the architecture, which is not recorded for this model — the numbers are indicative rather than exact.

What went into building it

The training run consumed about 4.8 × 10²⁰ FLOP, on NVIDIA V100. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

The training set ran to roughly 33,847,900,000 tokens.

It is tracked in the underlying dataset for one reason in particular: highly cited.

Step by step

How to choose a GPU for ESM1-670M (UR50/D)

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

  1. 01

    Check what it needs before anything else

    Look at what ESM1-670M (UR50/D) actually needs — around 1.4 GB at Q8_0. No amount of processing power compensates for a card that cannot hold it.

  2. 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 ESM1-670M (UR50/D).

  3. 03

    Set a quality floor

    Each card runs the least-compressed copy it can hold — Q8_0 on the smallest card that fits. Setting a floor drops the cards that only manage ESM1-670M (UR50/D) by squeezing it further than you would want.

  4. 04

    Rank by throughput rather than spec sheet

    Ranking by tokens per second for ESM1-670M (UR50/D) follows memory bandwidth, not core counts, which is why the B200 tops it at 5,063 tok/s.

  5. 05

    Look at the headroom, not just the fit

    The fit column separates cards that just manage ESM1-670M (UR50/D) from those with room to spare. Buy for the second if the context might grow.

  6. 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 ESM1-670M (UR50/D) alone — a card is usually bought for more than one model.

Answers

ESM1-670M (UR50/D) — common questions

01

Where can I download ESM1-670M (UR50/D)?

The weights for ESM1-670M (UR50/D) are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

02

How much compute was used to train ESM1-670M (UR50/D)?

Around 4.8 × 10²⁰ 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.

03

Can I run ESM1-670M (UR50/D) if it does not fit in my GPU?

It can be split between the card and system memory, but ESM1-670M (UR50/D) generates painfully slowly that way. Nothing on this page assumes offloading.

04

Would two GPUs run ESM1-670M (UR50/D) faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold ESM1-670M (UR50/D) on their own, a second card is rarely the answer here.

05

Why does the quantisation differ between cards for ESM1-670M (UR50/D)?

Because capacity varies, so does how hard ESM1-670M (UR50/D) has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

06

How accurate are these ESM1-670M (UR50/D) speed estimates?

These are estimates with real error bars. The fastest result here, 3,038–8,101 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

07

What GPU do I need to run ESM1-670M (UR50/D)?

The smallest card in our catalogue that holds ESM1-670M (UR50/D) is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.4 GB, and produces roughly 55.1 tokens per second. 818 cards in total can run it.

08

How fast is ESM1-670M (UR50/D) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 5,063 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 ESM1-670M (UR50/D) clear that.

09

How much VRAM does ESM1-670M (UR50/D) 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.

10

Can I run ESM1-670M (UR50/D) 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 943 tokens per second — a comfortable fit.

11

Can I run ESM1-670M (UR50/D) 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 577 tokens per second — a comfortable fit.

12

Can I run ESM1-670M (UR50/D) 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 715 tokens per second — a comfortable fit.

13

Can I run ESM1-670M (UR50/D) 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 848 tokens per second — a comfortable fit.

14

Is ESM1-670M (UR50/D) open source?

Its weights are published, so ESM1-670M (UR50/D) 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.

15

How many parameters does ESM1-670M (UR50/D) have?

ESM1-670M (UR50/D) has 669.2M parameters. See Table 1. 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.

16

Who created ESM1-670M (UR50/D)?

ESM1-670M (UR50/D) was published by Facebook AI Research,New York University (NYU), based in United States of America, categorised as industry,Academia.

17

When was ESM1-670M (UR50/D) released?

ESM1-670M (UR50/D) was published in August 2020. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

18

What is ESM1-670M (UR50/D) used for?

ESM1-670M (UR50/D) works in Biology, and is recorded as handling proteins, Protein or nucleotide language model (pLM/nLM). These are the areas it was designed around; they describe intent rather than a hard boundary.

Source

Original publication

Record last updated 11 February 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