Falcon-40B TPS calculator

Open weights Technology Innovation Institute 40B parameters March 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

126 of 818 cards that can run it

Smallest card that fits

Tesla M40 24 GB

24 GB · Q3_K_M · 7.0 tok/s

Fastest card

B200

84.7 tok/s · 180 GB

Which GPUs can run Falcon-40B?

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

126 cards match

Calculating

					Needs	Quantisation	Fit
84.7 tok/s 51–136 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	43.5 GB	Q8_0	Comfortable
84.7 tok/s 51–136 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	43.5 GB	Q8_0	Comfortable
67.6 tok/s 41–108 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	43.5 GB	Q8_0	Comfortable
67.6 tok/s 41–108 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	43.5 GB	Q8_0	Comfortable
54.1 tok/s 32–87 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	43.5 GB	Q8_0	Comfortable
51.8 tok/s 31–83 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	43.5 GB	Q8_0	Comfortable
51.8 tok/s 31–83 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	43.5 GB	Q8_0	Comfortable
49.6 tok/s 30–79 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	43.5 GB	Q8_0	Comfortable
45.7 tok/s 27–73 · low confidence	DRIVE A100 PROD NVIDIA	32 GB	1,870 GB/s	May 2020	24.9 GB	Q4_K_M	Tight
45.7 tok/s 27–73 · low confidence	GRID A100A NVIDIA	32 GB	1,870 GB/s	May 2020	24.9 GB	Q4_K_M	Tight
44.0 tok/s 26–70 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	43.5 GB	Q8_0	Comfortable
44.0 tok/s 26–70 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	43.5 GB	Q8_0	Comfortable
44.0 tok/s 26–70 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	43.5 GB	Q8_0	Comfortable
43.8 tok/s 26–70 · low confidence	GeForce RTX 5090 NVIDIA	32 GB	1,790 GB/s	Jan 2025	24.9 GB	Q4_K_M	Tight
43.8 tok/s 26–70 · low confidence	GeForce RTX 5090 D NVIDIA	32 GB	1,790 GB/s	Jan 2025	24.9 GB	Q4_K_M	Tight
41.7 tok/s 25–67 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	43.5 GB	Q8_0	Comfortable
38.3 tok/s 23–61 · low confidence	GeForce RTX 5090 D V2 NVIDIA	24 GB	1,340 GB/s	Aug 2025	20.2 GB	Q3_K_M	Tight
35.6 tok/s 21–57 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	43.5 GB	Q8_0	Comfortable
35.6 tok/s 21–57 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	43.5 GB	Q8_0	Comfortable
35.6 tok/s 21–57 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	43.5 GB	Q8_0	Comfortable
35.6 tok/s 21–57 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	43.5 GB	Q8_0	Comfortable
35.6 tok/s 21–57 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	43.5 GB	Q8_0	Comfortable
34.9 tok/s 21–56 · low confidence	A30X NVIDIA	24 GB	1,220 GB/s	Apr 2021	20.2 GB	Q3_K_M	Tight
28.9 tok/s 17–46 · low confidence	GeForce RTX 3090 Ti NVIDIA	24 GB	1,010 GB/s	Jan 2022	20.2 GB	Q3_K_M	Tight
28.9 tok/s 17–46 · low confidence	GeForce RTX 4090 NVIDIA	24 GB	1,010 GB/s	Sep 2022	20.2 GB	Q3_K_M	Tight

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: Technology Innovation Institute
Organisation type: Government
Country: United Arab Emirates
Published: 15 March 2023

What it does

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

Domain: Language
Task: Language modeling, 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: 40B
Training data: 1,000,000,000,000 tokens
Batch size: 2,359,296

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: 2.4 × 10²³ FLOP
How it was established: Operation counting,Reported

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
Chips used: 384
Chip-hours: 552,960
Wall-clock time: 1,440 hours (60 days)
Hardware utilisation: MFU 38.6%
Power draw: 306.3 kW
Compute cost: $319,783

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.

Foundation model: Yes
Likely above 10²³ FLOP: Yes
Why it is tracked: Historical significance
Record confidence: Confident

Sources

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

Reference: Abu Dhabi-based Technology Innovation Institute Introduces Falcon LLM: Foundational Large Language Model (LLM) outperforms GPT-3 with 40 Billion Parameters
Last updated: 28 November 2025

The extremes

The ten fastest GPUs for Falcon-40B

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 Falcon-40B

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

Hardware requirements in practice

Minimum card

Tesla M40 24 GB

Memory needed

20.2 GB

Fastest

84.7 tok/s

With 40B parameters, Falcon-40B lands in the range a serious desktop card can handle once the weights are compressed. 126 of the cards we track can run it.

The entry point is the Tesla M40 24 GB: 24 GB of memory, Q3_K_M compression, roughly 7.0 tokens per second.

At the other end, a B200 generates roughly 84.7 tokens per second on it, on the strength of 8,000 GB/s of memory bandwidth.

What this model is

Falcon-40B was published by Technology Innovation Institute, in United Arab Emirates, in March 2023. The organisation is categorised as government.

It works in Language, and is recorded as doing language modeling, 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.

What decides the speed

Across every card that can run it, the middle of the range is about 21.2 tokens per second, and 106 of them clear the ten tokens per second that roughly matches reading speed.

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.

Training and provenance

Training it took roughly 2.4 × 10²³ FLOP of computation, on NVIDIA A100 — a measure of what producing the model cost, not of how fast it answers.

The training set ran to roughly 1,000,000,000,000 tokens.

Its inclusion criterion is historical significance.

Step by step

How to choose a GPU for Falcon-40B

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 Falcon-40B — around 20.2 GB at Q3_K_M. Capacity is the gate — a card either holds it or it does not.
02

Decide how long your conversations run

The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context Falcon-40B can slip off a card that handles short questions easily.
03

Decide how much compression you will accept

The quantisation column varies by card, because a bigger card holds a more accurate copy of Falcon-40B — Q3_K_M on the smallest card that fits. Set a floor to hold the comparison at one level.
04

Compare tokens per second, not specifications

Ranking by tokens per second for Falcon-40B follows memory bandwidth, not core counts, which is why the B200 tops it at 84.7 tok/s.
05

Check the fit verdict before buying

Tight means Falcon-40B 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 Falcon-40B is settled.

Answers

Falcon-40B — common questions

What GPU do I need to run Falcon-40B?

The smallest card in our catalogue that holds Falcon-40B is the Tesla M40 24 GB, with 24 GB of memory. It runs the model at Q3_K_M using about 20.2 GB, and produces roughly 7.0 tokens per second. 126 cards in total can run it.

How fast is Falcon-40B on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 84.7 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 106 of the cards that can run Falcon-40B clear that.

How much VRAM does Falcon-40B need?

About 20.2 GB at Q3_K_M 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 Falcon-40B on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q3_K_M, using about 20.2 GB and generating roughly 38.3 tokens per second — a tight fit.

Is Falcon-40B open source?

Its weights are published, so Falcon-40B 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 Falcon-40B have?

Falcon-40B has 40B parameters. Model comes in 7B and 40B variants. 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 Falcon-40B?

Falcon-40B was published by Technology Innovation Institute, based in United Arab Emirates, categorised as government.

When was Falcon-40B released?

Falcon-40B was published in March 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.

What is Falcon-40B used for?

Falcon-40B works in Language, and is recorded as handling language modeling, Language modeling/generation, Question answering. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download Falcon-40B?

The weights for Falcon-40B 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 Falcon-40B?

Around 2.4 × 10²³ FLOP, on NVIDIA A100. 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 Falcon-40B 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 — the nearest miss we calculate is short by 6.9 GB. Our figures for Falcon-40B assume it is fully resident.

Would two GPUs run Falcon-40B faster?

Two cards buy memory rather than speed. That matters for Falcon-40B only if one card cannot hold it — 126 can, so a second adds little.

Why does the quantisation differ between cards for Falcon-40B?

Each card is shown running the least-compressed copy it can hold, and Falcon-40B appears at 4 different compression levels across the cards that fit it. Bigger cards get the more accurate version.

How accurate are these Falcon-40B speed estimates?

These are estimates with real error bars. The fastest result here, 51–136 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.

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