MatterSim (M3GNet - MatterSim-v1.0.0-5M) TPS calculator

Open weights Microsoft Research AI for Science 4.5M parameters May 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 · 8,192 tok/s

Fastest card

B200

752,941 tok/s · 180 GB

Which GPUs can run MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

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
752,941 tok/s

451,765–1,204,706 · low confidence

B200 NVIDIA 180 GB 8,000 GB/s Jan 2024 0.7 GB Q8_0 Comfortable
752,941 tok/s

451,765–1,204,706 · low confidence

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 0.7 GB Q8_0 Comfortable
601,242 tok/s

360,745–961,988 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 0.7 GB Q8_0 Comfortable
601,242 tok/s

360,745–961,988 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 0.7 GB Q8_0 Comfortable
480,847 tok/s

288,508–769,355 · low confidence

Radeon Instinct MI300 AMD 128 GB 6,550 GB/s Jan 2023 0.7 GB Q8_0 Comfortable
460,235 tok/s

276,141–736,376 · low confidence

H200 NVL NVIDIA 141 GB 4,890 GB/s Nov 2024 0.7 GB Q8_0 Comfortable
460,235 tok/s

276,141–736,376 · low confidence

H200 SXM 141 GB NVIDIA 141 GB 4,890 GB/s Nov 2024 0.7 GB Q8_0 Comfortable
440,471 tok/s

264,282–704,753 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 0.7 GB Q8_0 Comfortable
390,918 tok/s

234,551–625,468 · low confidence

Radeon Instinct MI300A AMD 128 GB 5,325 GB/s Dec 2023 0.7 GB Q8_0 Comfortable
390,918 tok/s

234,551–625,468 · low confidence

Radeon Instinct MI300X AMD 192 GB 5,325 GB/s Dec 2023 0.7 GB Q8_0 Comfortable
390,918 tok/s

234,551–625,468 · low confidence

Radeon Instinct MI308X AMD 192 GB 5,325 GB/s Dec 2023 0.7 GB Q8_0 Comfortable
370,824 tok/s

222,494–593,318 · low confidence

H100 NVL 94 GB NVIDIA 94 GB 3,940 GB/s Mar 2023 0.7 GB Q8_0 Comfortable
316,235 tok/s

189,741–505,976 · low confidence

H100 PCIe 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 0.7 GB Q8_0 Comfortable
316,235 tok/s

189,741–505,976 · low confidence

H100 SXM5 80 GB NVIDIA 80 GB 3,360 GB/s Oct 2022 0.7 GB Q8_0 Comfortable
316,235 tok/s

189,741–505,976 · low confidence

H100 SXM5 94 GB NVIDIA 94 GB 3,360 GB/s Mar 2023 0.7 GB Q8_0 Comfortable
316,235 tok/s

189,741–505,976 · low confidence

H100 SXM5 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 0.7 GB Q8_0 Comfortable
316,235 tok/s

189,741–505,976 · low confidence

H800 SXM5 NVIDIA 80 GB 3,360 GB/s Mar 2023 0.7 GB Q8_0 Comfortable
240,791 tok/s

144,474–385,265 · low confidence

Radeon Instinct MI250 AMD 128 GB 3,280 GB/s Nov 2021 0.7 GB Q8_0 Comfortable
240,791 tok/s

144,474–385,265 · low confidence

Radeon Instinct MI250X AMD 128 GB 3,280 GB/s Nov 2021 0.7 GB Q8_0 Comfortable
200,659 tok/s

120,395–321,054 · low confidence

Data Center GPU Max 1550 Intel 128 GB 3,280 GB/s Jan 2023 0.7 GB Q8_0 Comfortable
196,376 tok/s

117,826–314,202 · low confidence

Data Center GPU Max Subsystem Intel 128 GB 3,210 GB/s Jan 2023 0.7 GB Q8_0 Comfortable
192,000 tok/s

115,200–307,200 · low confidence

A100 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Nov 2020 0.7 GB Q8_0 Comfortable
192,000 tok/s

115,200–307,200 · low confidence

A100X NVIDIA 80 GB 2,040 GB/s Jun 2021 0.7 GB Q8_0 Comfortable
192,000 tok/s

115,200–307,200 · low confidence

A800 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Aug 2022 0.7 GB Q8_0 Comfortable
192,000 tok/s

115,200–307,200 · 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
Microsoft Research AI for Science
Organisation type
Industry
Country
United States of America
Published
10 May 2024
Authors
Han Yang, Chenxi Hu, Yichi Zhou, Xixian Liu, Yu Shi, Jielan Li, Guanzhi Li, Zekun Chen, Shuizhou Chen, Claudio Zeni, Matthew Horton, Robert Pinsler, Andrew Fowler, Daniel Zügner, Tian Xie, Jake Smith, Lixin Sun, Qian Wang, Lingyu Kong, Chang Liu, Hongxia Hao, Ziheng Lu

What it does

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

Domain
Materials science
Task
Atomistic simulations, Molecular simulation

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
4.5M

"As the training data size increases up to 3M, the the total number of parameters in M3GNet increase accordingly from 880K to 4.5M."

Training data
tokens

"As the training data size increases up to 3M, the the total number of parameters in M3GNet increase accordingly from 880K to 4.5M."

Epochs
200

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
1.6 × 10¹⁶ FLOP

Speculative confidence because I am unsure how to calculate gradient updates/tokens 6ND = 6*4500000*3000000*200 = 1.62e+16

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 A100
Chips used
8
Power draw
6.3 kW

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)

How it is classified

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

Record confidence
Speculative

Sources

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

Reference
MatterSim: A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures
Last updated
11 February 2026

The extremes

What the numbers mean

What it takes to run this model

Minimum card

Tesla C1080

Memory needed

0.7 GB

Fastest

752,941 tok/s

MatterSim (M3GNet - MatterSim-v1.0.0-5M) is small enough at 4.5M 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 8,192 tokens per second.

Top of the range is the B200, at roughly 752,941 tokens per second thanks to 8,000 GB/s of bandwidth.

What this model is

MatterSim (M3GNet - MatterSim-v1.0.0-5M) was published by Microsoft Research AI for Science, in United States of America, in May 2024. It comes out of industry.

It works in Materials science, and is recorded as doing atomistic simulations, Molecular simulation.

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

Half the cards that hold it manage more than 21,142.6 tokens per second, and 818 exceed reading speed outright.

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.

Without the attention layout on record, the memory column is an approximation. It is close enough to choose hardware by, and least reliable at long context.

Training and provenance

Training it took roughly 1.6 × 10¹⁶ FLOP of computation, on NVIDIA A100 — a measure of what producing the model cost, not of how fast it answers.

Step by step

How to choose a GPU for MatterSim (M3GNet - MatterSim-v1.0.0-5M)

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

  1. 01

    Read the memory figure first

    Every card here has been checked against MatterSim (M3GNet - MatterSim-v1.0.0-5M) — around 0.7 GB at Q8_0. Capacity is the gate — a card either holds it or it does not.

  2. 02

    Set the context length you will work at

    Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason MatterSim (M3GNet - MatterSim-v1.0.0-5M) stops fitting a card that seemed fine.

  3. 03

    Choose how far you will compress it

    Compression is what makes MatterSim (M3GNet - MatterSim-v1.0.0-5M) fit smaller cards, at some cost in accuracy — Q8_0 on the smallest card that fits. A minimum quality removes the ones that go too far.

  4. 04

    Rank by throughput rather than spec sheet

    Sort by speed to see how cards rank for MatterSim (M3GNet - MatterSim-v1.0.0-5M). It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 752,941 tok/s.

  5. 05

    Check the fit verdict before buying

    Tight means MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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.

  6. 06

    See what else that card runs

    Following a card through to its own page shows every other model it can hold, which is the question that follows once MatterSim (M3GNet - MatterSim-v1.0.0-5M) is settled.

Answers

MatterSim (M3GNet - MatterSim-v1.0.0-5M) — common questions

01

How much VRAM does MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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.

02

Can I run MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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 140,235 tokens per second — a comfortable fit.

03

Can I run MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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 85,873 tokens per second — a comfortable fit.

04

Can I run MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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 106,353 tokens per second — a comfortable fit.

05

Can I run MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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 126,118 tokens per second — a comfortable fit.

06

Is MatterSim (M3GNet - MatterSim-v1.0.0-5M) open source?

Its weights are published, so MatterSim (M3GNet - MatterSim-v1.0.0-5M) 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.

07

How many parameters does MatterSim (M3GNet - MatterSim-v1.0.0-5M) have?

MatterSim (M3GNet - MatterSim-v1.0.0-5M) has 4.5M parameters. "As the training data size increases up to 3M, the the total number of parameters in M3GNet increase accordingly from 880K to 4.5M.". 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.

08

Who created MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

MatterSim (M3GNet - MatterSim-v1.0.0-5M) was published by Microsoft Research AI for Science, based in United States of America, categorised as industry.

09

When was MatterSim (M3GNet - MatterSim-v1.0.0-5M) released?

MatterSim (M3GNet - MatterSim-v1.0.0-5M) was published in May 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.

10

What is MatterSim (M3GNet - MatterSim-v1.0.0-5M) used for?

MatterSim (M3GNet - MatterSim-v1.0.0-5M) works in Materials science, and is recorded as handling atomistic simulations, Molecular simulation. 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.

11

Where can I download MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

The weights for MatterSim (M3GNet - MatterSim-v1.0.0-5M) are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

12

How much compute was used to train MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

Around 1.6 × 10¹⁶ FLOP, on NVIDIA A100. 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.

13

Can I run MatterSim (M3GNet - MatterSim-v1.0.0-5M) if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part in system memory drags the whole thing down. Our figures for MatterSim (M3GNet - MatterSim-v1.0.0-5M) assume it is fully resident.

14

Would two GPUs run MatterSim (M3GNet - MatterSim-v1.0.0-5M) faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold MatterSim (M3GNet - MatterSim-v1.0.0-5M) on their own, a second card is rarely the answer here.

15

Why does the quantisation differ between cards for MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

Each card is shown running the least-compressed copy it can hold, and MatterSim (M3GNet - MatterSim-v1.0.0-5M) appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

16

How accurate are these MatterSim (M3GNet - MatterSim-v1.0.0-5M) speed estimates?

Every figure is derived from memory bandwidth and model size, not benchmarked. That is why each is published as a range such as 451,765–1,204,706 tok/s on the B200 rather than a single number.

17

What GPU do I need to run MatterSim (M3GNet - MatterSim-v1.0.0-5M)?

The smallest card in our catalogue that holds MatterSim (M3GNet - MatterSim-v1.0.0-5M) is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 0.7 GB, and produces roughly 8,192 tokens per second. 818 cards in total can run it.

18

How fast is MatterSim (M3GNet - MatterSim-v1.0.0-5M) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 752,941 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 MatterSim (M3GNet - MatterSim-v1.0.0-5M) clear that.

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.