ULM-FiT TPS calculator

Open weights University of San Francisco,Insight Centre NUI Galway,Fast.ai 441M parameters January 2018

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 · 83.6 tok/s

Fastest card

B200

7,683 tok/s · 180 GB

Which GPUs can run ULM-FiT?

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
7,683 tok/s 4,610–12,293 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	1.2 GB	Q8_0	Comfortable
7,683 tok/s 4,610–12,293 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	1.2 GB	Q8_0	Comfortable
6,135 tok/s 3,681–9,816 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	1.2 GB	Q8_0	Comfortable
6,135 tok/s 3,681–9,816 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	1.2 GB	Q8_0	Comfortable
4,907 tok/s 2,944–7,851 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	1.2 GB	Q8_0	Comfortable
4,696 tok/s 2,818–7,514 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.2 GB	Q8_0	Comfortable
4,696 tok/s 2,818–7,514 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	1.2 GB	Q8_0	Comfortable
4,495 tok/s 2,697–7,191 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	1.2 GB	Q8_0	Comfortable
3,989 tok/s 2,393–6,382 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	1.2 GB	Q8_0	Comfortable
3,989 tok/s 2,393–6,382 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	1.2 GB	Q8_0	Comfortable
3,989 tok/s 2,393–6,382 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	1.2 GB	Q8_0	Comfortable
3,784 tok/s 2,270–6,054 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	1.2 GB	Q8_0	Comfortable
3,227 tok/s 1,936–5,163 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.2 GB	Q8_0	Comfortable
3,227 tok/s 1,936–5,163 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	1.2 GB	Q8_0	Comfortable
3,227 tok/s 1,936–5,163 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	1.2 GB	Q8_0	Comfortable
3,227 tok/s 1,936–5,163 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	1.2 GB	Q8_0	Comfortable
3,227 tok/s 1,936–5,163 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	1.2 GB	Q8_0	Comfortable
2,457 tok/s 1,474–3,931 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	1.2 GB	Q8_0	Comfortable
2,457 tok/s 1,474–3,931 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	1.2 GB	Q8_0	Comfortable
2,048 tok/s 1,229–3,276 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	1.2 GB	Q8_0	Comfortable
2,004 tok/s 1,202–3,206 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	1.2 GB	Q8_0	Comfortable
1,959 tok/s 1,176–3,135 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	1.2 GB	Q8_0	Comfortable
1,959 tok/s 1,176–3,135 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	1.2 GB	Q8_0	Comfortable
1,959 tok/s 1,176–3,135 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	1.2 GB	Q8_0	Comfortable
1,959 tok/s 1,176–3,135 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	1.2 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: University of San Francisco,Insight Centre NUI Galway,Fast.ai
Organisation type: Academia,Academia
Country: United States of America, Ireland
Published: 18 January 2018
Authors: Jeremy Howard, Sebastian Ruder

What it does

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

Domain: Language
Task: Text classification
Base model: AWD-LSTM

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: 441M
Training data: 103,000,000 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: 2.7 × 10¹⁷ FLOP
How it was established: Operation counting

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.

Record confidence: Speculative
Citations: 1,940

Sources

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

Reference: Universal Language Model Fine-tuning for Text Classification
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run ULM-FiT

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 ULM-FiT

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

1.2 GB

Fastest

7,683 tok/s

ULM-FiT is small enough at 441M 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 83.6 tokens per second.

Top of the range is the B200, at roughly 7,683 tokens per second thanks to 8,000 GB/s of bandwidth.

About this model

ULM-FiT was published by University of San Francisco,Insight Centre NUI Galway,Fast.ai, in United States of America, in January 2018. It comes out of academia,Academia.

It works in Language, and is recorded as doing text classification.

It builds on AWD-LSTM, which is why it shares that model's general shape and size.

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.

How fast it runs, and why

The median result is around 215.7 tokens per second; 817 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.

Training and provenance

Training it took roughly 2.7 × 10¹⁷ FLOP of computation — a measure of what producing the model cost, not of how fast it answers.

Around 103,000,000 tokens went into training it.

Step by step

How to choose a GPU for ULM-FiT

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 ULM-FiT — around 1.2 GB at Q8_0. That figure, not the card's headline performance, is what decides whether it runs.
02

Match the context to your actual use

Longer conversations cost memory on top of what the weights need. Move the slider to your real working length before trusting any row for ULM-FiT.
03

Set a quality floor

Compression is what makes ULM-FiT 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

The speed ordering for ULM-FiT is effectively an ordering by memory bandwidth, which is why the B200 tops it at 7,683 tok/s.
05

Check the fit verdict before buying

Tight means ULM-FiT 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.
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 ULM-FiT is settled.

Answers

ULM-FiT — common questions

Where can I download ULM-FiT?

The weights for ULM-FiT 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 ULM-FiT?

Around 2.7 × 10¹⁷ FLOP. 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 ULM-FiT 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 ULM-FiT is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run ULM-FiT faster?

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

Why does the quantisation differ between cards for ULM-FiT?

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

How accurate are these ULM-FiT speed estimates?

These are estimates with real error bars. The fastest result here, 4,610–12,293 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

What GPU do I need to run ULM-FiT?

The smallest card in our catalogue that holds ULM-FiT is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 1.2 GB, and produces roughly 83.6 tokens per second. 818 cards in total can run it.

How fast is ULM-FiT on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 7,683 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 817 of the cards that can run ULM-FiT clear that.

How much VRAM does ULM-FiT need?

About 1.2 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 ULM-FiT on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 1.2 GB and generating roughly 1,431 tokens per second — a comfortable fit.

Can I run ULM-FiT on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 1.2 GB and generating roughly 876 tokens per second — a comfortable fit.

Can I run ULM-FiT on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 1.2 GB and generating roughly 1,085 tokens per second — a comfortable fit.

Can I run ULM-FiT on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 1.2 GB and generating roughly 1,287 tokens per second — a comfortable fit.

Is ULM-FiT open source?

Its weights are published, so ULM-FiT 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 ULM-FiT have?

ULM-FiT has 441M parameters. https://files.fast.ai/models/wt103/?C=S;O=D. 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 ULM-FiT?

ULM-FiT was published by University of San Francisco,Insight Centre NUI Galway,Fast.ai, based in United States of America, categorised as academia,Academia.

When was ULM-FiT released?

ULM-FiT was published in January 2018. 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 ULM-FiT used for?

ULM-FiT works in Language, and is recorded as handling text classification. These are the areas it was designed around; they describe intent rather than a hard boundary.

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