OpenELM-3B TPS calculator

Open weights Apple 3B 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 · Q6_K · 17.6 tok/s

Fastest card

B200

1,115 tok/s · 180 GB

Which GPUs can run OpenELM-3B?

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
1,115 tok/s 669–1,783 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	4.0 GB	Q8_0	Comfortable
1,115 tok/s 669–1,783 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	4.0 GB	Q8_0	Comfortable
890 tok/s 534–1,424 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	4.0 GB	Q8_0	Comfortable
890 tok/s 534–1,424 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	4.0 GB	Q8_0	Comfortable
712 tok/s 427–1,139 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	4.0 GB	Q8_0	Comfortable
681 tok/s 409–1,090 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	4.0 GB	Q8_0	Comfortable
681 tok/s 409–1,090 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	4.0 GB	Q8_0	Comfortable
652 tok/s 391–1,043 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	4.0 GB	Q8_0	Comfortable
579 tok/s 347–926 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	4.0 GB	Q8_0	Comfortable
579 tok/s 347–926 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	4.0 GB	Q8_0	Comfortable
579 tok/s 347–926 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	4.0 GB	Q8_0	Comfortable
549 tok/s 329–878 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	4.0 GB	Q8_0	Comfortable
468 tok/s 281–749 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	4.0 GB	Q8_0	Comfortable
468 tok/s 281–749 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	4.0 GB	Q8_0	Comfortable
468 tok/s 281–749 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	4.0 GB	Q8_0	Comfortable
468 tok/s 281–749 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	4.0 GB	Q8_0	Comfortable
468 tok/s 281–749 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	4.0 GB	Q8_0	Comfortable
356 tok/s 214–570 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	4.0 GB	Q8_0	Comfortable
356 tok/s 214–570 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	4.0 GB	Q8_0	Comfortable
297 tok/s 178–475 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	4.0 GB	Q8_0	Comfortable
291 tok/s 174–465 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	4.0 GB	Q8_0	Comfortable
284 tok/s 171–455 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	4.0 GB	Q8_0	Comfortable
284 tok/s 171–455 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	4.0 GB	Q8_0	Comfortable
284 tok/s 171–455 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	4.0 GB	Q8_0	Comfortable
284 tok/s 171–455 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	4.0 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: Apple
Organisation type: Industry
Country: United States of America
Published: 2 May 2024
Authors: Sachin Mehta, Mohammad Hossein Sekhavat, Qingqing Cao, Maxwell Horton, Yanzi Jin, Chenfan Sun, Iman Mirzadeh, Mahyar Najibi, Dmitry Belenko, Peter Zatloukal, Mohammad Rastegari

What it does

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

Domain: Language
Task: Language modeling/generation, Code generation, Question answering

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: 3B
Training data: tokens
Epochs: 1
Batch size: 4,000,000

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: 3.4 × 10²² FLOP
How it was established: Operation counting,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 H100 SXM5 80GB
Chips used: 128
Wall-clock time: 288 hours (12 days)
Power draw: 177.1 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 (non-commercial)
Hugging Face: apple

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: OpenELM: An Efficient Language Model Family with Open Training and Inference Framework
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run OpenELM-3B

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 OpenELM-3B

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 it takes to run this model

Minimum card

Tesla C1080

Memory needed

3.2 GB

Fastest

1,115 tok/s

OpenELM-3B is small enough at 3B 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 Q6_K and producing around 17.6 tokens per second.

Top of the range is the B200, at roughly 1,115 tokens per second thanks to 8,000 GB/s of bandwidth.

What this model is

OpenELM-3B was published by Apple, in United States of America, in May 2024. It comes out of industry.

It works in Language, and is recorded as doing language modeling/generation, Code generation, Question answering.

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. It is published under the apple organisation on Hugging Face.

What decides the speed

The median result is around 35.4 tokens per second; 780 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.

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

The training run consumed about 3.4 × 10²² FLOP, on NVIDIA H100 SXM5 80GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

Step by step

How to choose a GPU for OpenELM-3B

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

01

Start from the memory column

Every card here has been checked against OpenELM-3B — around 3.2 GB at Q6_K. 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 OpenELM-3B.
03

Decide how much compression you will accept

Compression is what makes OpenELM-3B fit smaller cards, at some cost in accuracy — Q6_K on the smallest card that fits. A minimum quality removes the ones that go too far.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for OpenELM-3B. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 1,115 tok/s.
05

Read the fit column last

The fit column separates cards that just manage OpenELM-3B from those with room to spare. Buy for the second if the context might grow.
06

See what else that card runs

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for OpenELM-3B alone — a card is usually bought for more than one model.

Answers

OpenELM-3B — common questions

Would two GPUs run OpenELM-3B faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run OpenELM-3B alone, the case for pairing is weak.

Why does the quantisation differ between cards for OpenELM-3B?

A larger card holds a more accurate copy. Across the cards that run OpenELM-3B, 2 compression levels are used; the floor control above pins it to one.

How accurate are these OpenELM-3B speed estimates?

They are calculated from specifications rather than measured, and each carries a range — 669–1,783 tok/s on the B200, for instance. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

What GPU do I need to run OpenELM-3B?

The smallest card in our catalogue that holds OpenELM-3B is the Tesla C1080, with 4 GB of memory. It runs the model at Q6_K using about 3.2 GB, and produces roughly 17.6 tokens per second. 818 cards in total can run it.

How fast is OpenELM-3B on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,115 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 780 of the cards that can run OpenELM-3B clear that.

How much VRAM does OpenELM-3B need?

About 3.2 GB at Q6_K 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 OpenELM-3B on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 4.0 GB and generating roughly 208 tokens per second — a comfortable fit.

Can I run OpenELM-3B on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 4.0 GB and generating roughly 127 tokens per second — a comfortable fit.

Can I run OpenELM-3B on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 4.0 GB and generating roughly 157 tokens per second — a comfortable fit.

Can I run OpenELM-3B on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 4.0 GB and generating roughly 187 tokens per second — a comfortable fit.

Is OpenELM-3B open source?

Its weights are published, so OpenELM-3B 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 OpenELM-3B have?

OpenELM-3B has 3B parameters. 3.04B (Table 4a). 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 OpenELM-3B?

OpenELM-3B was published by Apple, based in United States of America, categorised as industry.

When was OpenELM-3B released?

OpenELM-3B 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.

What is OpenELM-3B used for?

OpenELM-3B works in Language, and is recorded as handling language modeling/generation, Code generation, Question answering. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download OpenELM-3B?

Its weights are published under the apple organisation on Hugging Face. We do not host model files — this site calculates what hardware is needed to run them.

How much compute was used to train OpenELM-3B?

Around 3.4 × 10²² FLOP, on NVIDIA H100 SXM5 80GB. 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 OpenELM-3B 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 OpenELM-3B is rarely worth using. Every figure here assumes the whole model is on the card.

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