Ring-flash-linear-2.0 TPS calculator

Open weights Ant Group 104.2B parameters October 2025

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

43 cards that can run it

818 cards we hold specifications for

Smallest card that fits

Radeon Instinct MI200

64 GB · Q3_K_M · 14.0 tok/s

Fastest card

B200

32.5 tok/s · 180 GB

Which GPUs can run Ring-flash-linear-2.0?

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

43 cards match

Calculating

					Needs	Quantisation	Fit
32.5 tok/s 20–52 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	112.3 GB	Q8_0	Comfortable
32.5 tok/s 20–52 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	112.3 GB	Q8_0	Comfortable
31.5 tok/s 19–50 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	63.7 GB	Q4_K_M	Tight
31.5 tok/s 19–50 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	63.7 GB	Q4_K_M	Tight
28.6 tok/s 17–46 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	75.9 GB	Q5_K_M	Tight
26.0 tok/s 16–42 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	112.3 GB	Q8_0	Comfortable
26.0 tok/s 16–42 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	112.3 GB	Q8_0	Comfortable
24.4 tok/s 15–39 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	75.9 GB	Q5_K_M	Tight
24.4 tok/s 15–39 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	75.9 GB	Q5_K_M	Tight
24.4 tok/s 15–39 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	75.9 GB	Q5_K_M	Tight
22.2 tok/s 13–35 · low confidence	H100 SXM5 64 GB NVIDIA	64 GB	2,020 GB/s	Mar 2023	51.6 GB	Q3_K_M	Tight
20.8 tok/s 12–33 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	112.3 GB	Q8_0	Tight
19.9 tok/s 12–32 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	112.3 GB	Q8_0	Tight
19.9 tok/s 12–32 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	112.3 GB	Q8_0	Tight
19.1 tok/s 11–31 · low confidence	A100 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Nov 2020	63.7 GB	Q4_K_M	Tight
19.1 tok/s 11–31 · low confidence	A100X NVIDIA	80 GB	2,040 GB/s	Jun 2021	63.7 GB	Q4_K_M	Tight
19.1 tok/s 11–31 · low confidence	A800 SXM4 80 GB NVIDIA	80 GB	2,040 GB/s	Aug 2022	63.7 GB	Q4_K_M	Tight
19.1 tok/s 11–31 · low confidence	H100 CNX NVIDIA	80 GB	2,040 GB/s	Mar 2023	63.7 GB	Q4_K_M	Tight
19.1 tok/s 11–31 · low confidence	H100 PCIe 80 GB NVIDIA	80 GB	2,040 GB/s	Oct 2022	63.7 GB	Q4_K_M	Tight
19.1 tok/s 11–31 · low confidence	H800 PCIe 80 GB NVIDIA	80 GB	2,040 GB/s	Mar 2023	63.7 GB	Q4_K_M	Tight
19.0 tok/s 11–30 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	112.3 GB	Q8_0	Comfortable
18.2 tok/s 11–29 · low confidence	A100 PCIe 80 GB NVIDIA	80 GB	1,940 GB/s	Jun 2021	63.7 GB	Q4_K_M	Tight
18.2 tok/s 11–29 · low confidence	A800 PCIe 80 GB NVIDIA	80 GB	1,940 GB/s	Nov 2022	63.7 GB	Q4_K_M	Tight
16.9 tok/s 10–27 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	112.3 GB	Q8_0	Tight
16.9 tok/s 10–27 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	112.3 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: Ant Group
Organisation type: Industry
Country: China
Published: 23 October 2025
Authors: Bin Han, Caizhi Tang, Chen Liang, Donghao Zhang, Fan Yuan, Feng Zhu, Jie Gao, Jingyu Hu, Longfei Li, Meng Li, Mingyang Zhang, Peijie Jiang, Peng Jiao, Qian Zhao, Qingyuan Yang, Wenbo Shen, Xinxing Yang, Yalin Zhang, Yankun Ren, Yao Zhao, Yibo Cao, Yixuan Sun, Yue Zhang, Yuchen Fang, Zibin Lin, Zixuan Cheng, Jun Zhou

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, Quantitative reasoning, Code generation, Mathematical reasoning
Base model: Ling-flash-base-2.0-20T
Numerical format: FP8

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: 104.2B
Training data: 1,000,000,000,000 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: 7.7 × 10²³ FLOP
How it was established: Operation counting
Fine-tuning compute: 3.7 × 10²² 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 H800 SXM5
Chips used: 32
Power draw: 43.7 kW

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
Hugging Face: inclusionAI

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: Every Attention Matters: An Efficient Hybrid Architecture for Long-Context Reasoning
Last updated: 28 November 2025

The extremes

The ten fastest GPUs that run Ring-flash-linear-2.0

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 Ring-flash-linear-2.0

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 you need to run it

Minimum card

Radeon Instinct MI200

Memory needed

51.6 GB

Fastest

32.5 tok/s

Ring-flash-linear-2.0 sits at 104.2B parameters, which puts it above consumer hardware and into the range where a card is bought for this purpose rather than repurposed for it. 43 of the cards we track can hold it.

The least hardware that works is a Radeon Instinct MI200. Its 64 GB is enough at Q3_K_M compression, giving roughly 14.0 tokens per second.

The quickest result comes from a B200 at around 32.5 tokens per second — its 8,000 GB/s of bandwidth is what buys that.

Where it came from

Ring-flash-linear-2.0 was published by Ant Group, in China, in October 2025. The organisation is categorised as industry.

It works in Language, and is recorded as doing language modeling/generation, Question answering, Quantitative reasoning, Code generation, Mathematical reasoning.

It is derived from Ling-flash-base-2.0-20T rather than trained from scratch, which is the usual way a specialised model is produced.

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. It is published under the inclusionAI organisation on Hugging Face.

Understanding the speeds

The median result is around 18.2 tokens per second; 37 cards produce text faster than most people read it.

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

Training it took roughly 7.7 × 10²³ FLOP of computation, on NVIDIA H800 SXM5 — a measure of what producing the model cost, not of how fast it answers.

It was trained on about 1,000,000,000,000 tokens of text.

Step by step

How to choose a GPU for Ring-flash-linear-2.0

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 Ring-flash-linear-2.0 — around 51.6 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

The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context Ring-flash-linear-2.0 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 — Q3_K_M on the smallest card that fits. Setting a floor drops the cards that only manage Ring-flash-linear-2.0 by squeezing it further than you would want.
04

Rank by throughput rather than spec sheet

The speed ordering for Ring-flash-linear-2.0 is effectively an ordering by memory bandwidth, which is why the B200 tops it at 32.5 tok/s.
05

Check the fit verdict before buying

Tight means Ring-flash-linear-2.0 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

Open the card you have settled on

Following a card through to its own page shows every other model it can hold, which is the question that follows once Ring-flash-linear-2.0 is settled.

Answers

Ring-flash-linear-2.0 — common questions

What GPU do I need to run Ring-flash-linear-2.0?

The smallest card in our catalogue that holds Ring-flash-linear-2.0 is the Radeon Instinct MI200, with 64 GB of memory. It runs the model at Q3_K_M using about 51.6 GB, and produces roughly 14.0 tokens per second. 43 cards in total can run it.

How fast is Ring-flash-linear-2.0 on a GPU?

It depends on the card. The quickest we calculate is a B200 at about 32.5 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 37 of the cards that can run Ring-flash-linear-2.0 clear that.

How much VRAM does Ring-flash-linear-2.0 need?

About 51.6 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 Ring-flash-linear-2.0 open source?

Its weights are published, so Ring-flash-linear-2.0 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 Ring-flash-linear-2.0 have?

Ring-flash-linear-2.0 has 104.2B parameters. Ring-flash-linear-2.0 contains 104B parameters and 6.1B activations. 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 Ring-flash-linear-2.0?

Ring-flash-linear-2.0 was published by Ant Group, based in China, categorised as industry.

When was Ring-flash-linear-2.0 released?

Ring-flash-linear-2.0 was published in October 2025.

What is Ring-flash-linear-2.0 used for?

Ring-flash-linear-2.0 works in Language, and is recorded as handling language modeling/generation, Question answering, Quantitative reasoning, Code generation, Mathematical reasoning. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.

Where can I download Ring-flash-linear-2.0?

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

How much compute was used to train Ring-flash-linear-2.0?

Around 7.7 × 10²³ FLOP, on NVIDIA H800 SXM5. 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 Ring-flash-linear-2.0 if it does not fit in my GPU?

It can be split between the card and system memory, but Ring-flash-linear-2.0 generates painfully slowly that way — the nearest miss we calculate is short by 20.5 GB. Nothing on this page assumes offloading.

Would two GPUs run Ring-flash-linear-2.0 faster?

A second card roughly doubles the memory available but not the generation rate. With 43 cards already able to run Ring-flash-linear-2.0 alone, the case for pairing is weak.

Why does the quantisation differ between cards for Ring-flash-linear-2.0?

A larger card holds a more accurate copy. Across the cards that run Ring-flash-linear-2.0, 4 compression levels are used; the floor control above pins it to one.

How accurate are these Ring-flash-linear-2.0 speed estimates?

They are calculated from specifications rather than measured, and each carries a range — 20–52 tok/s on the B200, for instance. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

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