Hybrid H3-2.7B TPS calculator

Open weights Stanford University,University at Buffalo 2.7B parameters December 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 · 13.7 tok/s

Fastest card

B200

1,255 tok/s · 180 GB

Which GPUs can run Hybrid H3-2.7B?

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
1,255 tok/s 753–2,008 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	3.6 GB	Q8_0	Comfortable
1,255 tok/s 753–2,008 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	3.6 GB	Q8_0	Comfortable
1,002 tok/s 601–1,603 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	3.6 GB	Q8_0	Comfortable
1,002 tok/s 601–1,603 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	3.6 GB	Q8_0	Comfortable
801 tok/s 481–1,282 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	3.6 GB	Q8_0	Comfortable
767 tok/s 460–1,227 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.6 GB	Q8_0	Comfortable
767 tok/s 460–1,227 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.6 GB	Q8_0	Comfortable
734 tok/s 440–1,175 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	3.6 GB	Q8_0	Comfortable
652 tok/s 391–1,042 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	3.6 GB	Q8_0	Comfortable
652 tok/s 391–1,042 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	3.6 GB	Q8_0	Comfortable
652 tok/s 391–1,042 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	3.6 GB	Q8_0	Comfortable
618 tok/s 371–989 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	3.6 GB	Q8_0	Comfortable
527 tok/s 316–843 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.6 GB	Q8_0	Comfortable
527 tok/s 316–843 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	3.6 GB	Q8_0	Comfortable
527 tok/s 316–843 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	3.6 GB	Q8_0	Comfortable
527 tok/s 316–843 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.6 GB	Q8_0	Comfortable
527 tok/s 316–843 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	3.6 GB	Q8_0	Comfortable
401 tok/s 241–642 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	3.6 GB	Q8_0	Comfortable
401 tok/s 241–642 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	3.6 GB	Q8_0	Comfortable
334 tok/s 201–535 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	3.6 GB	Q8_0	Comfortable
327 tok/s 196–524 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	3.6 GB	Q8_0	Comfortable
320 tok/s 192–512 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	3.6 GB	Q8_0	Comfortable
320 tok/s 192–512 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	3.6 GB	Q8_0	Comfortable
320 tok/s 192–512 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	3.6 GB	Q8_0	Comfortable
320 tok/s 192–512 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	3.6 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: Stanford University,University at Buffalo
Organisation type: Academia,Academia
Country: United States of America
Published: 28 December 2022
Authors: Daniel Y. Fu, Tri Dao, Khaled K. Saab, Armin W. Thomas, Atri Rudra, Christopher Ré

What it does

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

Domain: Language
Task: Language modeling/generation, Question answering
Numerical format: BF16

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: 2.7B
Training data: 400,000,000,000 tokens
Epochs: 509.02
Batch size: 1,048,576

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: 6.5 × 10²¹ 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 A100 SXM4 80 GB
Chips used: 8
Power draw: 6.4 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)
Training code: Unreleased

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: SOTA improvement
Record confidence: Likely
Citations: 636
Benchmark data: Hybrid H3-2.7B

Sources

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

Reference: Hungry Hungry Hippos: Towards Language Modeling with State Space Models
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run Hybrid H3-2.7B

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 Hybrid H3-2.7B

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

3.6 GB

Fastest

1,255 tok/s

Hybrid H3-2.7B is small enough at 2.7B 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 13.7 tokens per second.

At the other end, a B200 generates roughly 1,255 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

Where it came from

Hybrid H3-2.7B was published by Stanford University,University at Buffalo, in United States of America, in December 2022. The organisation is categorised as academia,Academia.

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

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.

Understanding the speeds

The median result is around 35.2 tokens per second; 775 cards produce text faster than most people read it.

Every weight participates in every token here, so bandwidth is the whole story: the ranking below is effectively a ranking of memory throughput.

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

The training run consumed about 6.5 × 10²¹ FLOP, on NVIDIA A100 SXM4 80 GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

The training set ran to roughly 400,000,000,000 tokens.

Its inclusion criterion is sOTA improvement.

Step by step

How to choose a GPU for Hybrid H3-2.7B

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 Hybrid H3-2.7B — around 3.6 GB at Q8_0. That figure, not the card's headline performance, is what decides whether it runs.
02

Decide how long your conversations run

Set the context to what you will actually use. The cache grows with the conversation, and it is the usual reason Hybrid H3-2.7B stops fitting a card that seemed fine.
03

Decide how much compression you will accept

Compression is what makes Hybrid H3-2.7B 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.
04

Compare tokens per second, not specifications

Sort by speed to see how cards rank for Hybrid H3-2.7B. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 1,255 tok/s.
05

Read the fit column last

The fit column separates cards that just manage Hybrid H3-2.7B from those with room to spare. Buy for the second if the context might grow.
06

Open the card you have settled on

Each card page repeats this sweep for every model we hold. It answers what else the hardware is good for, beyond Hybrid H3-2.7B.

Answers

Hybrid H3-2.7B — common questions

Who created Hybrid H3-2.7B?

Hybrid H3-2.7B was published by Stanford University,University at Buffalo, based in United States of America, categorised as academia,Academia.

When was Hybrid H3-2.7B released?

Hybrid H3-2.7B was published in December 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 Hybrid H3-2.7B used for?

Hybrid H3-2.7B works in Language, and is recorded as handling language modeling/generation, Question answering. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download Hybrid H3-2.7B?

The weights for Hybrid H3-2.7B 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 Hybrid H3-2.7B?

Around 6.5 × 10²¹ FLOP, on NVIDIA A100 SXM4 80 GB. 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 Hybrid H3-2.7B if it does not fit in my GPU?

It can be split between the card and system memory, but Hybrid H3-2.7B generates painfully slowly that way. Nothing on this page assumes offloading.

Would two GPUs run Hybrid H3-2.7B faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold Hybrid H3-2.7B on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for Hybrid H3-2.7B?

A larger card holds a more accurate copy. Across the cards that run Hybrid H3-2.7B, 1 compression levels are used; the floor control above pins it to one.

How accurate are these Hybrid H3-2.7B 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 753–2,008 tok/s on the B200 rather than a single number.

What GPU do I need to run Hybrid H3-2.7B?

The smallest card in our catalogue that holds Hybrid H3-2.7B is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 3.6 GB, and produces roughly 13.7 tokens per second. 818 cards in total can run it.

How fast is Hybrid H3-2.7B on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,255 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 775 of the cards that can run Hybrid H3-2.7B clear that.

How much VRAM does Hybrid H3-2.7B need?

About 3.6 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 Hybrid H3-2.7B on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 3.6 GB and generating roughly 234 tokens per second — a comfortable fit.

Can I run Hybrid H3-2.7B on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 3.6 GB and generating roughly 143 tokens per second — a comfortable fit.

Can I run Hybrid H3-2.7B on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 3.6 GB and generating roughly 177 tokens per second — a comfortable fit.

Can I run Hybrid H3-2.7B on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 3.6 GB and generating roughly 210 tokens per second — a comfortable fit.

Is Hybrid H3-2.7B open source?

Its weights are published, so Hybrid H3-2.7B 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 Hybrid H3-2.7B have?

Hybrid H3-2.7B has 2.7B parameters. 2.7B. 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.

Source

Original publication

Record last updated 25 May 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