GemNet-T (OC20) TPS calculator

Open weights Technical University of Munich 1.9M parameters July 2021

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 · 19,402 tok/s

Fastest card

B200

1,783,282 tok/s · 180 GB

Which GPUs can run GemNet-T (OC20)?

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
1,783,282 tok/s 1,069,969–2,853,251 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	0.7 GB	Q8_0	Comfortable
1,783,282 tok/s 1,069,969–2,853,251 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	0.7 GB	Q8_0	Comfortable
1,423,995 tok/s 854,397–2,278,392 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
1,423,995 tok/s 854,397–2,278,392 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
1,138,848 tok/s 683,309–1,822,157 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
1,090,031 tok/s 654,019–1,744,050 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
1,090,031 tok/s 654,019–1,744,050 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
1,043,220 tok/s 625,932–1,669,152 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	0.7 GB	Q8_0	Comfortable
925,858 tok/s 555,515–1,481,372 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
925,858 tok/s 555,515–1,481,372 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
925,858 tok/s 555,515–1,481,372 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
878,266 tok/s 526,960–1,405,226 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
748,978 tok/s 449,387–1,198,365 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
748,978 tok/s 449,387–1,198,365 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	0.7 GB	Q8_0	Comfortable
748,978 tok/s 449,387–1,198,365 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
748,978 tok/s 449,387–1,198,365 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
748,978 tok/s 449,387–1,198,365 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
570,294 tok/s 342,176–912,470 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
570,294 tok/s 342,176–912,470 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
475,245 tok/s 285,147–760,391 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
465,102 tok/s 279,061–744,163 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
454,737 tok/s 272,842–727,579 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	0.7 GB	Q8_0	Comfortable
454,737 tok/s 272,842–727,579 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	0.7 GB	Q8_0	Comfortable
454,737 tok/s 272,842–727,579 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	0.7 GB	Q8_0	Comfortable
454,737 tok/s 272,842–727,579 · 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: Technical University of Munich
Organisation type: Academia
Country: Germany
Published: 1 July 2021
Authors: Johannes Gasteiger, Florian Becker, Stephan Günnemann

What it does

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

Domain: Materials science
Task: Molecular simulation, Molecular property prediction, Molecular representation learning, Atomistic simulations

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: 1.9M
Training data: 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: 7.3 × 10¹⁷ 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 GeForce GTX 1080 Ti
Chips used: 1
Wall-clock time: 60 hours
Power draw: 278 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 (restricted use)
Training code: Open (restricted use)

How it is classified

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

Record confidence: Likely

Sources

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

Reference: GemNet: Universal Directional Graph Neural Networks for Molecules
Last updated: 11 February 2026

The extremes

The ten fastest GPUs that run GemNet-T (OC20)

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 GemNet-T (OC20)

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

1,783,282 tok/s

GemNet-T (OC20) is small enough at 1.9M 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 19,402 tokens per second.

The quickest result comes from a B200 at around 1,783,282 tokens per second — its 8,000 GB/s of bandwidth is what buys that.

What this model is

GemNet-T (OC20) was published by Technical University of Munich, in Germany, in July 2021. academia is the category the publisher falls under.

It works in Materials science, and is recorded as doing molecular simulation, Molecular property prediction, Molecular representation learning, Atomistic simulations.

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

The median result is around 50,074.6 tokens per second; 818 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.

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.

What went into building it

Producing it required around 7.3 × 10¹⁷ FLOP of arithmetic, on NVIDIA GeForce GTX 1080 Ti, which is a statement about the training budget rather than about inference.

Step by step

How to choose a GPU for GemNet-T (OC20)

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 GemNet-T (OC20) — 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

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for GemNet-T (OC20).
03

Set a quality floor

The quantisation column varies by card, because a bigger card holds a more accurate copy of GemNet-T (OC20) — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for GemNet-T (OC20). It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 1,783,282 tok/s.
05

Look at the headroom, not just the fit

A tight fit runs GemNet-T (OC20) but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
06

See what else that card runs

Each card page repeats this sweep for every model we hold. It answers what else the hardware is good for, beyond GemNet-T (OC20).

Answers

GemNet-T (OC20) — common questions

What is GemNet-T (OC20) used for?

GemNet-T (OC20) works in Materials science, and is recorded as handling molecular simulation, Molecular property prediction, Molecular representation learning, Atomistic simulations. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download GemNet-T (OC20)?

The weights for GemNet-T (OC20) 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 GemNet-T (OC20)?

Around 7.3 × 10¹⁷ FLOP, on NVIDIA GeForce GTX 1080 Ti. 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 GemNet-T (OC20) 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 GemNet-T (OC20) assume it is fully resident.

Would two GPUs run GemNet-T (OC20) faster?

Two cards buy memory rather than speed. That matters for GemNet-T (OC20) only if one card cannot hold it — 818 can, so a second adds little.

Why does the quantisation differ between cards for GemNet-T (OC20)?

Each card is shown running the least-compressed copy it can hold, and GemNet-T (OC20) appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

How accurate are these GemNet-T (OC20) 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 1,069,969–2,853,251 tok/s on the B200 rather than a single number.

What GPU do I need to run GemNet-T (OC20)?

The smallest card in our catalogue that holds GemNet-T (OC20) is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 0.7 GB, and produces roughly 19,402 tokens per second. 818 cards in total can run it.

How fast is GemNet-T (OC20) on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,783,282 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 GemNet-T (OC20) clear that.

How much VRAM does GemNet-T (OC20) 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 GemNet-T (OC20) 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 332,136 tokens per second — a comfortable fit.

Can I run GemNet-T (OC20) 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 203,383 tokens per second — a comfortable fit.

Can I run GemNet-T (OC20) 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 251,889 tokens per second — a comfortable fit.

Can I run GemNet-T (OC20) 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 298,700 tokens per second — a comfortable fit.

Is GemNet-T (OC20) open source?

Its weights are published, so GemNet-T (OC20) 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 GemNet-T (OC20) have?

GemNet-T (OC20) has 1.9M parameters. 1.9M parameters. 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 GemNet-T (OC20)?

GemNet-T (OC20) was published by Technical University of Munich, based in Germany, categorised as academia.

When was GemNet-T (OC20) released?

GemNet-T (OC20) was published in July 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

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