OpenCLIP ViT-H-14-378-quickgelu TPS calculator

Open weights Massachusetts Institute of Technology (MIT) 986.7M parameters April 2023

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 · 37.4 tok/s

Fastest card

B200

3,434 tok/s · 180 GB

Which GPUs can run OpenCLIP ViT-H-14-378-quickgelu?

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
3,434 tok/s

2,060–5,494 · low confidence

B200 NVIDIA 180 GB 8,000 GB/s Jan 2024 1.8 GB Q8_0 Comfortable
3,434 tok/s

2,060–5,494 · low confidence

B300 NVIDIA 288 GB 8,000 GB/s Sep 2025 1.8 GB Q8_0 Comfortable
2,742 tok/s

1,645–4,387 · low confidence

Radeon Instinct MI350X AMD 288 GB 8,190 GB/s Jan 2025 1.8 GB Q8_0 Comfortable
2,742 tok/s

1,645–4,387 · low confidence

Radeon Instinct MI355X AMD 288 GB 8,190 GB/s Jan 2025 1.8 GB Q8_0 Comfortable
2,193 tok/s

1,316–3,509 · low confidence

Radeon Instinct MI300 AMD 128 GB 6,550 GB/s Jan 2023 1.8 GB Q8_0 Comfortable
2,099 tok/s

1,259–3,358 · low confidence

H200 NVL NVIDIA 141 GB 4,890 GB/s Nov 2024 1.8 GB Q8_0 Comfortable
2,099 tok/s

1,259–3,358 · low confidence

H200 SXM 141 GB NVIDIA 141 GB 4,890 GB/s Nov 2024 1.8 GB Q8_0 Comfortable
2,009 tok/s

1,205–3,214 · low confidence

Radeon Instinct MI325X AMD 256 GB 6,000 GB/s Oct 2024 1.8 GB Q8_0 Comfortable
1,783 tok/s

1,070–2,853 · low confidence

Radeon Instinct MI300A AMD 128 GB 5,325 GB/s Dec 2023 1.8 GB Q8_0 Comfortable
1,783 tok/s

1,070–2,853 · low confidence

Radeon Instinct MI300X AMD 192 GB 5,325 GB/s Dec 2023 1.8 GB Q8_0 Comfortable
1,783 tok/s

1,070–2,853 · low confidence

Radeon Instinct MI308X AMD 192 GB 5,325 GB/s Dec 2023 1.8 GB Q8_0 Comfortable
1,691 tok/s

1,015–2,706 · low confidence

H100 NVL 94 GB NVIDIA 94 GB 3,940 GB/s Mar 2023 1.8 GB Q8_0 Comfortable
1,442 tok/s

865–2,308 · low confidence

H100 PCIe 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 1.8 GB Q8_0 Comfortable
1,442 tok/s

865–2,308 · low confidence

H100 SXM5 80 GB NVIDIA 80 GB 3,360 GB/s Oct 2022 1.8 GB Q8_0 Comfortable
1,442 tok/s

865–2,308 · low confidence

H100 SXM5 94 GB NVIDIA 94 GB 3,360 GB/s Mar 2023 1.8 GB Q8_0 Comfortable
1,442 tok/s

865–2,308 · low confidence

H100 SXM5 96 GB NVIDIA 96 GB 3,360 GB/s Mar 2023 1.8 GB Q8_0 Comfortable
1,442 tok/s

865–2,308 · low confidence

H800 SXM5 NVIDIA 80 GB 3,360 GB/s Mar 2023 1.8 GB Q8_0 Comfortable
1,098 tok/s

659–1,757 · low confidence

Radeon Instinct MI250 AMD 128 GB 3,280 GB/s Nov 2021 1.8 GB Q8_0 Comfortable
1,098 tok/s

659–1,757 · low confidence

Radeon Instinct MI250X AMD 128 GB 3,280 GB/s Nov 2021 1.8 GB Q8_0 Comfortable
915 tok/s

549–1,464 · low confidence

Data Center GPU Max 1550 Intel 128 GB 3,280 GB/s Jan 2023 1.8 GB Q8_0 Comfortable
896 tok/s

537–1,433 · low confidence

Data Center GPU Max Subsystem Intel 128 GB 3,210 GB/s Jan 2023 1.8 GB Q8_0 Comfortable
876 tok/s

525–1,401 · low confidence

A100 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Nov 2020 1.8 GB Q8_0 Comfortable
876 tok/s

525–1,401 · low confidence

A100X NVIDIA 80 GB 2,040 GB/s Jun 2021 1.8 GB Q8_0 Comfortable
876 tok/s

525–1,401 · low confidence

A800 SXM4 80 GB NVIDIA 80 GB 2,040 GB/s Aug 2022 1.8 GB Q8_0 Comfortable
876 tok/s

525–1,401 · low confidence

H100 CNX NVIDIA 80 GB 2,040 GB/s Mar 2023 1.8 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
Massachusetts Institute of Technology (MIT)
Organisation type
Academia
Country
United States of America
Published
16 April 2023
Authors
Ross Wightman, Romain Beaumont, Cade Gordon, Vaishaal Shankar

What it does

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

Domain
Vision
Task
Image captioning

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
986.7M

See row 2 of the csv here: https://github.com/mlfoundations/open_clip/blob/main/docs/openclip_results.csv

Training data
tokens

The model is said to have seen 44B samples (https://github.com/mlfoundations/open_clip?tab=readme-ov-file#openclip).

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)

https://github.com/mlfoundations/open_clip custom license

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
OpenCLIP
Last updated
28 November 2025

The extremes

The ten fastest GPUs that run OpenCLIP ViT-H-14-378-quickgelu

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 3,434 tok/s
  2. 02 B200 180 GB · 8,000 GB/s · Q8_0 3,434 tok/s
  3. 03 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q8_0 2,742 tok/s
  4. 04 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q8_0 2,742 tok/s
  5. 05 Radeon Instinct MI300 128 GB · 6,550 GB/s · Q8_0 2,193 tok/s
  6. 06 H200 NVL 141 GB · 4,890 GB/s · Q8_0 2,099 tok/s
  7. 07 H200 SXM 141 GB 141 GB · 4,890 GB/s · Q8_0 2,099 tok/s
  8. 08 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q8_0 2,009 tok/s
  9. 09 Radeon Instinct MI300A 128 GB · 5,325 GB/s · Q8_0 1,783 tok/s
  10. 10 Radeon Instinct MI300X 192 GB · 5,325 GB/s · Q8_0 1,783 tok/s

The smallest GPUs that still run OpenCLIP ViT-H-14-378-quickgelu

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 1.8 GB · Q8_0 · comfortable 41.2 tok/s
  2. 02 RTX A400 4 GB · needs 1.8 GB · Q8_0 · comfortable 41.2 tok/s
  3. 03 RTX 500 Mobile Ada Generation 4 GB · needs 1.8 GB · Q8_0 · comfortable 54.9 tok/s
  4. 04 GeForce RTX 3050 A Mobile 4 GB · needs 1.8 GB · Q8_0 · comfortable 82.4 tok/s
  5. 05 Jetson Orin Nano 4 GB 4 GB · needs 1.8 GB · Q8_0 · comfortable 14.6 tok/s
  6. 06 Radeon RX 6450M 4 GB · needs 1.8 GB · Q8_0 · comfortable 42.9 tok/s
  7. 07 Radeon RX 6550M 4 GB · needs 1.8 GB · Q8_0 · comfortable 48.2 tok/s
  8. 08 Radeon RX 6550S 4 GB · needs 1.8 GB · Q8_0 · comfortable 42.9 tok/s
  9. 09 Arc A310 4 GB · needs 1.8 GB · Q8_0 · comfortable 34.6 tok/s
  10. 10 Arc Pro A30M 4 GB · needs 1.8 GB · Q8_0 · comfortable 35.7 tok/s

What the numbers mean

The hardware side

Minimum card

Tesla C1080

Memory needed

1.8 GB

Fastest

3,434 tok/s

OpenCLIP ViT-H-14-378-quickgelu is small enough at 986.7M 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 37.4 tokens per second.

The quickest result comes from a B200 at around 3,434 tokens per second — its 8,000 GB/s of bandwidth is what buys that.

Where it came from

OpenCLIP ViT-H-14-378-quickgelu was published by Massachusetts Institute of Technology (MIT), in United States of America, in April 2023. The organisation is categorised as academia.

It works in Vision, and is recorded as doing image captioning.

The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached.

Understanding the speeds

The median result is around 96.4 tokens per second; 806 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.

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.

Step by step

How to choose a GPU for OpenCLIP ViT-H-14-378-quickgelu

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

  1. 01

    Start from the memory column

    Look at what OpenCLIP ViT-H-14-378-quickgelu actually needs — around 1.8 GB at Q8_0. No amount of processing power compensates for a card that cannot hold it.

  2. 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 OpenCLIP ViT-H-14-378-quickgelu stops fitting a card that seemed fine.

  3. 03

    Set a quality floor

    The quantisation column varies by card, because a bigger card holds a more accurate copy of OpenCLIP ViT-H-14-378-quickgelu — Q8_0 on the smallest card that fits. Set a floor to hold the comparison at one level.

  4. 04

    Rank by throughput rather than spec sheet

    Sort by speed to see how cards rank for OpenCLIP ViT-H-14-378-quickgelu. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 3,434 tok/s.

  5. 05

    Look at the headroom, not just the fit

    A tight fit runs OpenCLIP ViT-H-14-378-quickgelu but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.

  6. 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 OpenCLIP ViT-H-14-378-quickgelu alone — a card is usually bought for more than one model.

Answers

OpenCLIP ViT-H-14-378-quickgelu — common questions

01

What is OpenCLIP ViT-H-14-378-quickgelu used for?

OpenCLIP ViT-H-14-378-quickgelu works in Vision, and is recorded as handling image captioning. These are the areas it was designed around; they describe intent rather than a hard boundary.

02

Where can I download OpenCLIP ViT-H-14-378-quickgelu?

The weights for OpenCLIP ViT-H-14-378-quickgelu are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

03

Can I run OpenCLIP ViT-H-14-378-quickgelu 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 OpenCLIP ViT-H-14-378-quickgelu is rarely worth using. Every figure here assumes the whole model is on the card.

04

Would two GPUs run OpenCLIP ViT-H-14-378-quickgelu faster?

Two cards buy memory rather than speed. That matters for OpenCLIP ViT-H-14-378-quickgelu only if one card cannot hold it — 818 can, so a second adds little.

05

Why does the quantisation differ between cards for OpenCLIP ViT-H-14-378-quickgelu?

Each card is shown running the least-compressed copy it can hold, and OpenCLIP ViT-H-14-378-quickgelu appears at 1 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

06

How accurate are these OpenCLIP ViT-H-14-378-quickgelu 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 2,060–5,494 tok/s on the B200 rather than a single number.

07

What GPU do I need to run OpenCLIP ViT-H-14-378-quickgelu?

The smallest card in our catalogue that holds OpenCLIP ViT-H-14-378-quickgelu is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.8 GB, and produces roughly 37.4 tokens per second. 818 cards in total can run it.

08

How fast is OpenCLIP ViT-H-14-378-quickgelu on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 3,434 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 806 of the cards that can run OpenCLIP ViT-H-14-378-quickgelu clear that.

09

How much VRAM does OpenCLIP ViT-H-14-378-quickgelu need?

About 1.8 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 OpenCLIP ViT-H-14-378-quickgelu on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 1.8 GB and generating roughly 640 tokens per second — a comfortable fit.

11

Can I run OpenCLIP ViT-H-14-378-quickgelu on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 1.8 GB and generating roughly 392 tokens per second — a comfortable fit.

12

Can I run OpenCLIP ViT-H-14-378-quickgelu on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 1.8 GB and generating roughly 485 tokens per second — a comfortable fit.

13

Can I run OpenCLIP ViT-H-14-378-quickgelu on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 1.8 GB and generating roughly 575 tokens per second — a comfortable fit.

14

Is OpenCLIP ViT-H-14-378-quickgelu open source?

Its weights are published, so OpenCLIP ViT-H-14-378-quickgelu 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 OpenCLIP ViT-H-14-378-quickgelu have?

OpenCLIP ViT-H-14-378-quickgelu has 986.7M parameters. See row 2 of the csv here: https://github.com/mlfoundations/open_clip/blob/main/docs/openclip_results.csv. 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 OpenCLIP ViT-H-14-378-quickgelu?

OpenCLIP ViT-H-14-378-quickgelu was published by Massachusetts Institute of Technology (MIT), based in United States of America, categorised as academia.

17

When was OpenCLIP ViT-H-14-378-quickgelu released?

OpenCLIP ViT-H-14-378-quickgelu was published in April 2023. 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 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