Theseus 6/768 TPS calculator

Open weights University of California San Diego,Beihang University,Microsoft 66M parameters February 2020

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 · 559 tok/s

Fastest card

B200

51,337 tok/s · 180 GB

Which GPUs can run Theseus 6/768?

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
51,337 tok/s 30,802–82,139 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	0.8 GB	Q8_0	Comfortable
51,337 tok/s 30,802–82,139 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	0.8 GB	Q8_0	Comfortable
40,994 tok/s 24,596–65,590 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	0.8 GB	Q8_0	Comfortable
40,994 tok/s 24,596–65,590 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	0.8 GB	Q8_0	Comfortable
32,785 tok/s 19,671–52,456 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	0.8 GB	Q8_0	Comfortable
31,380 tok/s 18,828–50,207 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.8 GB	Q8_0	Comfortable
31,380 tok/s 18,828–50,207 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.8 GB	Q8_0	Comfortable
30,032 tok/s 18,019–48,051 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	0.8 GB	Q8_0	Comfortable
26,653 tok/s 15,992–42,646 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	0.8 GB	Q8_0	Comfortable
26,653 tok/s 15,992–42,646 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	0.8 GB	Q8_0	Comfortable
26,653 tok/s 15,992–42,646 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	0.8 GB	Q8_0	Comfortable
25,283 tok/s 15,170–40,453 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	0.8 GB	Q8_0	Comfortable
21,562 tok/s 12,937–34,498 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.8 GB	Q8_0	Comfortable
21,562 tok/s 12,937–34,498 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	0.8 GB	Q8_0	Comfortable
21,562 tok/s 12,937–34,498 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	0.8 GB	Q8_0	Comfortable
21,562 tok/s 12,937–34,498 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.8 GB	Q8_0	Comfortable
21,562 tok/s 12,937–34,498 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	0.8 GB	Q8_0	Comfortable
16,418 tok/s 9,851–26,268 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	0.8 GB	Q8_0	Comfortable
16,418 tok/s 9,851–26,268 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	0.8 GB	Q8_0	Comfortable
13,681 tok/s 8,209–21,890 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	0.8 GB	Q8_0	Comfortable
13,389 tok/s 8,034–21,423 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	0.8 GB	Q8_0	Comfortable
13,091 tok/s 7,855–20,945 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	0.8 GB	Q8_0	Comfortable
13,091 tok/s 7,855–20,945 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	0.8 GB	Q8_0	Comfortable
13,091 tok/s 7,855–20,945 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	0.8 GB	Q8_0	Comfortable
13,091 tok/s 7,855–20,945 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	0.8 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: University of California San Diego,Beihang University,Microsoft
Organisation type: Academia,Academia,Industry
Country: United States of America, China
Published: 7 February 2020
Authors: Canwen Xu, Wangchunshu Zhou, Tao Ge, Furu Wei, Ming Zhou

What it does

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

Domain: Language
Task: Text autocompletion
Base model: BERT-Large

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: 66M
Training data: 393,000 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.

Fine-tuning compute: 2.7 × 10¹⁸ FLOP

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 V100

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.

Why it is tracked: SOTA improvement
Record confidence: Confident
Citations: 224

Sources

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

Reference: BERT-of-Theseus: Compressing BERT by Progressive Module Replacing
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run Theseus 6/768

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 Theseus 6/768

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 C1080

Memory needed

0.8 GB

Fastest

51,337 tok/s

Theseus 6/768 is small enough at 66M parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The entry point is the Tesla C1080: 4 GB of memory, Q8_0 compression, roughly 559 tokens per second.

A B200 is the fastest we calculate for it: about 51,337 tokens per second, from 8,000 GB/s of memory bandwidth.

Where it came from

Theseus 6/768 was published by University of California San Diego,Beihang University,Microsoft, in United States of America, in February 2020. The organisation is categorised as academia,Academia,Industry.

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

It is derived from BERT-Large rather than trained from scratch, which is the usual way a specialised model is produced.

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded.

Understanding the speeds

Across every card that can run it, the middle of the range is about 1,441.5 tokens per second, and 818 of them clear the ten tokens per second that roughly matches reading speed.

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.

How it was trained

The training set ran to roughly 393,000 tokens.

The reason it appears in this catalogue at all is sOTA improvement.

Step by step

How to choose a GPU for Theseus 6/768

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

01

Check what it needs before anything else

Look at what Theseus 6/768 actually needs — around 0.8 GB at Q8_0. No amount of processing power compensates for a card that cannot hold it.
02

Set the context length you will work at

The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context Theseus 6/768 can slip off a card that handles short questions easily.
03

Decide how much compression you will accept

Each card runs the least-compressed copy it can hold — Q8_0 on the smallest card that fits. Setting a floor drops the cards that only manage Theseus 6/768 by squeezing it further than you would want.
04

Rank by throughput rather than spec sheet

The speed ordering for Theseus 6/768 is effectively an ordering by memory bandwidth, which is why the B200 tops it at 51,337 tok/s.
05

Look at the headroom, not just the fit

The fit column separates cards that just manage Theseus 6/768 from those with room to spare. Buy for the second if the context might grow.
06

Check the card from the other side

Every card name links to its own page, which runs the same calculation across the whole model catalogue. Worth a look before buying for Theseus 6/768 alone — a card is usually bought for more than one model.

Answers

Theseus 6/768 — common questions

What GPU do I need to run Theseus 6/768?

The smallest card in our catalogue that holds Theseus 6/768 is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 0.8 GB, and produces roughly 559 tokens per second. 818 cards in total can run it.

How fast is Theseus 6/768 on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 51,337 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 818 of the cards that can run Theseus 6/768 clear that.

How much VRAM does Theseus 6/768 need?

About 0.8 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.

Can I run Theseus 6/768 on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 0.8 GB and generating roughly 9,562 tokens per second — a comfortable fit.

Can I run Theseus 6/768 on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 0.8 GB and generating roughly 5,855 tokens per second — a comfortable fit.

Can I run Theseus 6/768 on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 0.8 GB and generating roughly 7,251 tokens per second — a comfortable fit.

Can I run Theseus 6/768 on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 0.8 GB and generating roughly 8,599 tokens per second — a comfortable fit.

Is Theseus 6/768 open source?

Its weights are published, so Theseus 6/768 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 Theseus 6/768 have?

Theseus 6/768 has 66M parameters. 66M, Table 1. 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 Theseus 6/768?

Theseus 6/768 was published by University of California San Diego,Beihang University,Microsoft, based in United States of America, categorised as academia,Academia,Industry.

When was Theseus 6/768 released?

Theseus 6/768 was published in February 2020. 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 Theseus 6/768 used for?

Theseus 6/768 works in Language, and is recorded as handling text autocompletion. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download Theseus 6/768?

The weights for Theseus 6/768 are published, though we do not hold a repository link for it. This site calculates hardware requirements rather than hosting model files.

Can I run Theseus 6/768 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 Theseus 6/768 is rarely worth using. Every figure here assumes the whole model is on the card.

Would two GPUs run Theseus 6/768 faster?

A second card roughly doubles the memory available but not the generation rate. With 818 cards already able to run Theseus 6/768 alone, the case for pairing is weak.

Why does the quantisation differ between cards for Theseus 6/768?

A larger card holds a more accurate copy. Across the cards that run Theseus 6/768, 1 compression levels are used; the floor control above pins it to one.

How accurate are these Theseus 6/768 speed estimates?

These are estimates with real error bars. The fastest result here, 30,802–82,139 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 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