Pythia-2.8b TPS calculator

Open weights EleutherAI,Booz Allen Hamilton, McLean,University of Cambridge,Indraprastha Institute of Information Technology Delhi,Stability AI,datasaur.ai,University of Amsterdam 2.8B 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 · Q6_K · 19.1 tok/s

Fastest card

B200

1,210 tok/s · 180 GB

Which GPUs can run Pythia-2.8b?

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
1,210 tok/s 726–1,936 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	3.7 GB	Q8_0	Comfortable
1,210 tok/s 726–1,936 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	3.7 GB	Q8_0	Comfortable
966 tok/s 580–1,546 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	3.7 GB	Q8_0	Comfortable
966 tok/s 580–1,546 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	3.7 GB	Q8_0	Comfortable
773 tok/s 464–1,236 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	3.7 GB	Q8_0	Comfortable
740 tok/s 444–1,183 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.7 GB	Q8_0	Comfortable
740 tok/s 444–1,183 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	3.7 GB	Q8_0	Comfortable
708 tok/s 425–1,133 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	3.7 GB	Q8_0	Comfortable
628 tok/s 377–1,005 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	3.7 GB	Q8_0	Comfortable
628 tok/s 377–1,005 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	3.7 GB	Q8_0	Comfortable
628 tok/s 377–1,005 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	3.7 GB	Q8_0	Comfortable
596 tok/s 358–954 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	3.7 GB	Q8_0	Comfortable
508 tok/s 305–813 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.7 GB	Q8_0	Comfortable
508 tok/s 305–813 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	3.7 GB	Q8_0	Comfortable
508 tok/s 305–813 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	3.7 GB	Q8_0	Comfortable
508 tok/s 305–813 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	3.7 GB	Q8_0	Comfortable
508 tok/s 305–813 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	3.7 GB	Q8_0	Comfortable
387 tok/s 232–619 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	3.7 GB	Q8_0	Comfortable
387 tok/s 232–619 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	3.7 GB	Q8_0	Comfortable
322 tok/s 193–516 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	3.7 GB	Q8_0	Comfortable
316 tok/s 189–505 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	3.7 GB	Q8_0	Comfortable
309 tok/s 185–494 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	3.7 GB	Q8_0	Comfortable
309 tok/s 185–494 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	3.7 GB	Q8_0	Comfortable
309 tok/s 185–494 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	3.7 GB	Q8_0	Comfortable
309 tok/s 185–494 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	3.7 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: EleutherAI,Booz Allen Hamilton, McLean,University of Cambridge,Indraprastha Institute of Information Technology Delhi,Stability AI,datasaur.ai,University of Amsterdam
Organisation type: Research collective,Industry,Academia,Academia,Industry,Industry,Academia
Country: United States of America, United Kingdom of Great Britain and Northern Ireland, India, Netherlands
Published: 3 April 2023
Authors: Stella Biderman, Hailey Schoelkopf, Quentin Anthony, Herbie Bradley, Kyle O'Brien, Eric Hallahan, Mohammad Aflah Khan, Shivanshu Purohit, USVSN Sai Prashanth, Edward Raff, Aviya Skowron, Lintang Sutawika, Oskar van der Wal

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

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: 2.8B
Training data: 299,892,736,000 tokens
Epochs: 1

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: 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 40 GB
Chips used: 64
Chip-hours: 14,240
Hardware utilisation: MFU 31.5%
Power draw: 51.0 kW
Cloud vendor: Amazon Web Services
Data centre: AWS US East

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: Open source

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence: Confident
Citations: 1,862
Benchmark data: Pythia-2.8b

Sources

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

Reference: Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run Pythia-2.8b

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 Pythia-2.8b

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

The hardware side

Minimum card

Tesla C1080

Memory needed

3.0 GB

Fastest

1,210 tok/s

Pythia-2.8b is small enough at 2.8B 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 Q6_K, for about 19.1 tokens per second.

A B200 is the fastest we calculate for it: about 1,210 tokens per second, from 8,000 GB/s of memory bandwidth.

Where it came from

Pythia-2.8b was published by EleutherAI,Booz Allen Hamilton, McLean,University of Cambridge,Indraprastha Institute of Information Technology Delhi,Stability AI,datasaur.ai,University of Amsterdam, in United States of America, in April 2023. The organisation is categorised as research collective,Industry,Academia,Academia,Industry,Industry,Academia.

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

Understanding the speeds

Half the cards that hold it manage more than 38.4 tokens per second, and 784 exceed reading speed outright.

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.

How it was trained

Producing it required around 5 × 10²¹ FLOP of arithmetic, on NVIDIA A100 SXM4 40 GB, which is a statement about the training budget rather than about inference.

The training set ran to roughly 299,892,736,000 tokens.

Step by step

How to choose a GPU for Pythia-2.8b

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 Pythia-2.8b — around 3.0 GB at Q6_K. Capacity is the gate — a card either holds it or it does not.
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 Pythia-2.8b.
03

Choose how far you will compress it

The quantisation column varies by card, because a bigger card holds a more accurate copy of Pythia-2.8b — Q6_K 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 Pythia-2.8b follows memory bandwidth, not core counts, which is why the B200 tops it at 1,210 tok/s.
05

Check the fit verdict before buying

The fit column separates cards that just manage Pythia-2.8b 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 Pythia-2.8b alone — a card is usually bought for more than one model.

Answers

Pythia-2.8b — common questions

When was Pythia-2.8b released?

Pythia-2.8b 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.

What is Pythia-2.8b used for?

Pythia-2.8b works in Language, and is recorded as handling language modeling/generation, Question answering. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

Where can I download Pythia-2.8b?

The weights for Pythia-2.8b 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 Pythia-2.8b?

Around 5 × 10²¹ FLOP, on NVIDIA A100 SXM4 40 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 Pythia-2.8b if it does not fit in my GPU?

It can be split between the card and system memory, but Pythia-2.8b generates painfully slowly that way. Nothing on this page assumes offloading.

Would two GPUs run Pythia-2.8b faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run Pythia-2.8b alone, the case for pairing is weak.

Why does the quantisation differ between cards for Pythia-2.8b?

Because capacity varies, so does how hard Pythia-2.8b has to be squeezed — 2 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these Pythia-2.8b speed estimates?

These are estimates with real error bars. The fastest result here, 726–1,936 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 Pythia-2.8b?

The smallest card in our catalogue that holds Pythia-2.8b is the Tesla C1080, with 4 GB of memory. It runs the model at Q6_K using about 3.0 GB, and produces roughly 19.1 tokens per second. 818 cards in total can run it.

How fast is Pythia-2.8b on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,210 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 784 of the cards that can run Pythia-2.8b clear that.

How much VRAM does Pythia-2.8b need?

About 3.0 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 Pythia-2.8b on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 3.7 GB and generating roughly 225 tokens per second — a comfortable fit.

Can I run Pythia-2.8b on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 3.7 GB and generating roughly 138 tokens per second — a comfortable fit.

Can I run Pythia-2.8b on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 3.7 GB and generating roughly 171 tokens per second — a comfortable fit.

Can I run Pythia-2.8b on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 3.7 GB and generating roughly 203 tokens per second — a comfortable fit.

Is Pythia-2.8b open source?

Its weights are published, so Pythia-2.8b 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 Pythia-2.8b have?

Pythia-2.8b has 2.8B parameters. See Table 1 for non-embedding parameters. 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 Pythia-2.8b?

Pythia-2.8b was published by EleutherAI,Booz Allen Hamilton, McLean,University of Cambridge,Indraprastha Institute of Information Technology Delhi,Stability AI,datasaur.ai,University of Amsterdam, based in United States of America, categorised as research collective,Industry,Academia,Academia,Industry,Industry,Academia.

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