KataGo TPS calculator

Open weights Jane Street 2.5M parameters February 2019

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 · 14,746 tok/s

Fastest card

B200

1,355,294 tok/s · 180 GB

Which GPUs can run KataGo?

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
1,355,294 tok/s 813,176–2,168,471 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	0.7 GB	Q8_0	Comfortable
1,355,294 tok/s 813,176–2,168,471 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	0.7 GB	Q8_0	Comfortable
1,082,236 tok/s 649,342–1,731,578 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
1,082,236 tok/s 649,342–1,731,578 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	0.7 GB	Q8_0	Comfortable
865,525 tok/s 519,315–1,384,840 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
828,424 tok/s 497,054–1,325,478 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
828,424 tok/s 497,054–1,325,478 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	0.7 GB	Q8_0	Comfortable
792,847 tok/s 475,708–1,268,555 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	0.7 GB	Q8_0	Comfortable
703,652 tok/s 422,191–1,125,843 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
703,652 tok/s 422,191–1,125,843 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
703,652 tok/s 422,191–1,125,843 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	0.7 GB	Q8_0	Comfortable
667,482 tok/s 400,489–1,067,972 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
569,224 tok/s 341,534–910,758 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
569,224 tok/s 341,534–910,758 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	0.7 GB	Q8_0	Comfortable
569,224 tok/s 341,534–910,758 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
569,224 tok/s 341,534–910,758 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
569,224 tok/s 341,534–910,758 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	0.7 GB	Q8_0	Comfortable
433,423 tok/s 260,054–693,477 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
433,423 tok/s 260,054–693,477 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	0.7 GB	Q8_0	Comfortable
361,186 tok/s 216,712–577,897 · low confidence	Data Center GPU Max 1550 Intel	128 GB	3,280 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
353,478 tok/s 212,087–565,564 · low confidence	Data Center GPU Max Subsystem Intel	128 GB	3,210 GB/s	Jan 2023	0.7 GB	Q8_0	Comfortable
345,600 tok/s 207,360–552,960 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	0.7 GB	Q8_0	Comfortable
345,600 tok/s 207,360–552,960 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	0.7 GB	Q8_0	Comfortable
345,600 tok/s 207,360–552,960 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	0.7 GB	Q8_0	Comfortable
345,600 tok/s 207,360–552,960 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	0.7 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: Jane Street
Organisation type: Industry
Country: United States of America
Published: 27 February 2019
Authors: David J. Wu

What it does

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

Domain: Games
Task: Go
Approach: Self-supervised learning
Numerical format: FP16

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: 2.5M
Training data: 241,000,000 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: 2.3 × 10¹⁹ FLOP
How it was established: Hardware

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 Tesla V100 DGXS 16 GB
Wall-clock time: 456 hours (19 days)
Compute cost: $105

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: Speculative
Citations: 111

Sources

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

Reference: Accelerating Self-Play Learning in Go
Last updated: 25 May 2026

The extremes

The ten fastest GPUs that run KataGo

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 KataGo

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 it takes to run this model

Minimum card

Tesla C1080

Memory needed

0.7 GB

Fastest

1,355,294 tok/s

KataGo is small enough at 2.5M parameters that hardware is rarely the obstacle — 818 of the cards we track can run it, including cards several years old.

The smallest card that holds it is the Tesla C1080 with 4 GB, running it at Q8_0 and producing around 14,746 tokens per second.

Top of the range is the B200, at roughly 1,355,294 tokens per second thanks to 8,000 GB/s of bandwidth.

Where it came from

KataGo was published by Jane Street, in United States of America, in February 2019. industry is the category the publisher falls under.

It works in Games, and is recorded as doing go.

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 38,056.7 tokens per second; 818 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.

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

The training run consumed about 2.3 × 10¹⁹ FLOP, on NVIDIA Tesla V100 DGXS 16 GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.

Around 241,000,000 tokens went into training it.

Its inclusion criterion is sOTA improvement.

Step by step

How to choose a GPU for KataGo

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 KataGo — around 0.7 GB at Q8_0. That figure, not the card's headline performance, is what decides whether it runs.
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 KataGo can slip off a card that handles short questions easily.
03

Decide how much compression you will accept

Compression is what makes KataGo 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

Rank by throughput rather than spec sheet

Ranking by tokens per second for KataGo follows memory bandwidth, not core counts, which is why the B200 tops it at 1,355,294 tok/s.
05

Read the fit column last

A tight fit runs KataGo but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
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 KataGo alone — a card is usually bought for more than one model.

Answers

KataGo — common questions

Who created KataGo?

KataGo was published by Jane Street, based in United States of America, categorised as industry.

When was KataGo released?

KataGo was published in February 2019. 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 KataGo used for?

KataGo works in Games, and is recorded as handling go. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download KataGo?

The weights for KataGo 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 KataGo?

Around 2.3 × 10¹⁹ FLOP, on NVIDIA Tesla V100 DGXS 16 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 KataGo 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. Our figures for KataGo assume it is fully resident.

Would two GPUs run KataGo faster?

Capacity adds across cards; throughput does not. Since 818 of the cards we track already hold KataGo on their own, a second card is rarely the answer here.

Why does the quantisation differ between cards for KataGo?

Because capacity varies, so does how hard KataGo has to be squeezed — 1 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these KataGo speed estimates?

These are estimates with real error bars. The fastest result here, 813,176–2,168,471 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 KataGo?

The smallest card in our catalogue that holds KataGo is the Tesla C1080, with 4 GB of memory. It runs the model at Q8_0 using about 0.7 GB, and produces roughly 14,746 tokens per second. 818 cards in total can run it.

How fast is KataGo on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 1,355,294 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 KataGo clear that.

How much VRAM does KataGo need?

About 0.7 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 KataGo on a 8 GB GPU?

Yes. A CMP 170HX 8 GB with 8 GB runs it at Q8_0, using about 0.7 GB and generating roughly 252,424 tokens per second — a comfortable fit.

Can I run KataGo on a 12 GB GPU?

Yes. A GeForce RTX 3080 Ti with 12 GB runs it at Q8_0, using about 0.7 GB and generating roughly 154,571 tokens per second — a comfortable fit.

Can I run KataGo on a 16 GB GPU?

Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at Q8_0, using about 0.7 GB and generating roughly 191,435 tokens per second — a comfortable fit.

Can I run KataGo on a 24 GB GPU?

Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q8_0, using about 0.7 GB and generating roughly 227,012 tokens per second — a comfortable fit.

Is KataGo open source?

Its weights are published, so KataGo 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 KataGo have?

KataGo has 2.5M parameters. https://arxiv.org/abs/2210.00849 gives parameter count for AlphaZero in Fig 1b. 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.

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