APUS-xDAN-4.0(MoE) TPS calculator

Open weights Qilin Hesheng Network Technology Co., Ltd. (APUS) 136B parameters April 2024

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

37 of 818 cards that can run it

Smallest card that fits

A100 SXM4 80 GB

80 GB · Q3_K_M · 17.1 tok/s

Fastest card

H100 NVL 94 GB

28.3 tok/s · 94 GB

Which GPUs can run APUS-xDAN-4.0(MoE)?

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

37 cards match

Calculating

					Needs	Quantisation	Fit
28.3 tok/s 17–45 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	83.0 GB	Q4_K_M	Tight
28.2 tok/s 17–45 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	67.1 GB	Q3_K_M	Tight
28.2 tok/s 17–45 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	67.1 GB	Q3_K_M	Tight
24.9 tok/s 15–40 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	146.3 GB	Q8_0	Tight
24.9 tok/s 15–40 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	146.3 GB	Q8_0	Comfortable
24.2 tok/s 14–39 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	83.0 GB	Q4_K_M	Tight
24.2 tok/s 14–39 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	83.0 GB	Q4_K_M	Tight
24.2 tok/s 14–39 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	83.0 GB	Q4_K_M	Tight
23.1 tok/s 14–37 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	114.6 GB	Q6_K	Tight
22.1 tok/s 13–35 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	114.6 GB	Q6_K	Tight
22.1 tok/s 13–35 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	114.6 GB	Q6_K	Tight
19.9 tok/s 12–32 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	146.3 GB	Q8_0	Comfortable
19.9 tok/s 12–32 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	146.3 GB	Q8_0	Comfortable
18.8 tok/s 11–30 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	114.6 GB	Q6_K	Tight
17.1 tok/s 10–27 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	67.1 GB	Q3_K_M	Tight
17.1 tok/s 10–27 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	67.1 GB	Q3_K_M	Tight
17.1 tok/s 10–27 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	67.1 GB	Q3_K_M	Tight
17.1 tok/s 10–27 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	67.1 GB	Q3_K_M	Tight
17.1 tok/s 10–27 · low confidence	H100 PCIe 80 GB NVIDIA	80 GB	2,040 GB/s	Oct 2022	67.1 GB	Q3_K_M	Tight
17.1 tok/s 10–27 · low confidence	H800 PCIe 80 GB NVIDIA	80 GB	2,040 GB/s	Mar 2023	67.1 GB	Q3_K_M	Tight
16.3 tok/s 10–26 · low confidence	A100 PCIe 80 GB NVIDIA	80 GB	1,940 GB/s	Jun 2021	67.1 GB	Q3_K_M	Tight
16.3 tok/s 10–26 · low confidence	A800 PCIe 80 GB NVIDIA	80 GB	1,940 GB/s	Nov 2022	67.1 GB	Q3_K_M	Tight
14.6 tok/s 9–23 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	146.3 GB	Q8_0	Comfortable
12.9 tok/s 8–21 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	146.3 GB	Q8_0	Tight
12.9 tok/s 8–21 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	146.3 GB	Q8_0	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: Qilin Hesheng Network Technology Co., Ltd. (APUS)
Organisation type: Industry
Country: China
Published: 2 April 2024

What it does

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

Domain: Multimodal, Language
Task: Chat, Recommender system, Image generation, Language modeling/generation, Question answering, Mathematical reasoning, Quantitative reasoning

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: 136B
Training data: tokens

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 (restricted use)
Training code: Unreleased
Hugging Face: xDAN-AI

How it is classified

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

Record confidence: Confident

Sources

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

Reference: A Trillion-Parameter MOE Architecture Model Outperforms Grok1, Compatible with 4090 Graphics Card
Last updated: 11 February 2026

The extremes

The ten fastest GPUs for APUS-xDAN-4.0(MoE)

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 APUS-xDAN-4.0(MoE)

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

A100 SXM4 80 GB

Memory needed

67.1 GB

Fastest

28.3 tok/s

APUS-xDAN-4.0(MoE) sits at 136B parameters, which puts it above consumer hardware and into the range where a card is bought for this purpose rather than repurposed for it. 37 of the cards we track can hold it.

At the low end, a A100 SXM4 80 GB handles it — 80 GB, at Q3_K_M, for about 17.1 tokens per second.

The quickest result comes from a H100 NVL 94 GB at around 28.3 tokens per second — its 3,940 GB/s of bandwidth is what buys that.

About this model

APUS-xDAN-4.0(MoE) was published by Qilin Hesheng Network Technology Co., Ltd. (APUS), in China, in April 2024. The organisation is categorised as industry.

It works in Multimodal, Language, and is recorded as doing chat, Recommender system, Image generation, Language modeling/generation, Question answering, Mathematical reasoning, Quantitative reasoning.

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. It is published under the xDAN-AI organisation on Hugging Face.

How fast it runs, and why

Half the cards that hold it manage more than 17.1 tokens per second, and 33 exceed reading speed outright.

Every weight participates in every token here, so bandwidth is the whole story: the ranking below is effectively a ranking of memory throughput.

Memory here is estimated from size rather than computed from the architecture, which is not recorded for this model — the numbers are indicative rather than exact.

Step by step

How to choose a GPU for APUS-xDAN-4.0(MoE)

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

01

Start from the memory column

Every card here has been checked against APUS-xDAN-4.0(MoE) — around 67.1 GB at Q3_K_M. 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 APUS-xDAN-4.0(MoE).
03

Set a quality floor

Compression is what makes APUS-xDAN-4.0(MoE) fit smaller cards, at some cost in accuracy — Q3_K_M on the smallest card that fits. A minimum quality removes the ones that go too far.
04

Rank by throughput rather than spec sheet

Sort by speed to see how cards rank for APUS-xDAN-4.0(MoE). It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the H100 NVL 94 GB tops it at 28.3 tok/s.
05

Look at the headroom, not just the fit

A tight fit runs APUS-xDAN-4.0(MoE) 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 APUS-xDAN-4.0(MoE) alone — a card is usually bought for more than one model.

Answers

APUS-xDAN-4.0(MoE) — common questions

Would two GPUs run APUS-xDAN-4.0(MoE) faster?

A second card roughly doubles the memory available but not the generation rate. With 37 cards already able to run APUS-xDAN-4.0(MoE) alone, the case for pairing is weak.

Why does the quantisation differ between cards for APUS-xDAN-4.0(MoE)?

Because capacity varies, so does how hard APUS-xDAN-4.0(MoE) has to be squeezed — 5 distinct levels appear in the table above. Set a minimum quality to compare at one.

How accurate are these APUS-xDAN-4.0(MoE) speed estimates?

These are estimates with real error bars. The fastest result here, 17–45 tok/s on the H100 NVL 94 GB, could reasonably land anywhere in its published range depending on which runtime you use.

What GPU do I need to run APUS-xDAN-4.0(MoE)?

The smallest card in our catalogue that holds APUS-xDAN-4.0(MoE) is the A100 SXM4 80 GB, with 80 GB of memory. It runs the model at Q3_K_M using about 67.1 GB, and produces roughly 17.1 tokens per second. 37 cards in total can run it.

How fast is APUS-xDAN-4.0(MoE) on a GPU?

It depends on the card. The quickest we calculate is a H100 NVL 94 GB at about 28.3 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 33 of the cards that can run APUS-xDAN-4.0(MoE) clear that.

How much VRAM does APUS-xDAN-4.0(MoE) need?

About 67.1 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.

Is APUS-xDAN-4.0(MoE) open source?

Its weights are published, so APUS-xDAN-4.0(MoE) 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 APUS-xDAN-4.0(MoE) have?

APUS-xDAN-4.0(MoE) has 136B parameters. According to reports, APUS-xDAN-4.0 (MoE) is the first open source large model in China with a MoE architecture of more than 100 billion parameters, with a parameter scale of 136 billion. This is also the largest open source model in China. Among the previous large open source models in China, the largest parameter scale was Alibaba's Qianwen 72B with 72 billion parameters, with a parameter scale of 72 billion. 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 APUS-xDAN-4.0(MoE)?

APUS-xDAN-4.0(MoE) was published by Qilin Hesheng Network Technology Co., Ltd. (APUS), based in China, categorised as industry.

When was APUS-xDAN-4.0(MoE) released?

APUS-xDAN-4.0(MoE) was published in April 2024. 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 APUS-xDAN-4.0(MoE) used for?

APUS-xDAN-4.0(MoE) works in Multimodal, Language, and is recorded as handling chat, Recommender system, Image generation, Language modeling/generation, Question answering, Mathematical reasoning, Quantitative reasoning. These are the areas it was designed around; they describe intent rather than a hard boundary.

Where can I download APUS-xDAN-4.0(MoE)?

Its weights are published under the xDAN-AI organisation on Hugging Face. We do not host model files — this site calculates what hardware is needed to run them.

Can I run APUS-xDAN-4.0(MoE) 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 APUS-xDAN-4.0(MoE) is rarely worth using — the nearest miss we calculate is short by 18.2 GB. Every figure here assumes the whole model is on the card.

Source

Original publication

Record last updated 11 February 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