AgentFounder-30B TPS calculator

Open weights Alibaba 30B parameters September 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

132 cards that can run it

818 cards we hold specifications for

Smallest card that fits

RTX A4500

20 GB · IQ4_XS · 22.2 tok/s

Fastest card

B200

113 tok/s · 180 GB

Which GPUs can run AgentFounder-30B?

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

132 cards match

Calculating

					Needs	Quantisation	Fit
113 tok/s 68–181 · low confidence	B200 NVIDIA	180 GB	8,000 GB/s	Jan 2024	32.8 GB	Q8_0	Comfortable
113 tok/s 68–181 · low confidence	B300 NVIDIA	288 GB	8,000 GB/s	Sep 2025	32.8 GB	Q8_0	Comfortable
90.2 tok/s 54–144 · low confidence	Radeon Instinct MI350X AMD	288 GB	8,190 GB/s	Jan 2025	32.8 GB	Q8_0	Comfortable
90.2 tok/s 54–144 · low confidence	Radeon Instinct MI355X AMD	288 GB	8,190 GB/s	Jan 2025	32.8 GB	Q8_0	Comfortable
72.1 tok/s 43–115 · low confidence	Radeon Instinct MI300 AMD	128 GB	6,550 GB/s	Jan 2023	32.8 GB	Q8_0	Comfortable
69.0 tok/s 41–110 · low confidence	H200 NVL NVIDIA	141 GB	4,890 GB/s	Nov 2024	32.8 GB	Q8_0	Comfortable
69.0 tok/s 41–110 · low confidence	H200 SXM 141 GB NVIDIA	141 GB	4,890 GB/s	Nov 2024	32.8 GB	Q8_0	Comfortable
66.1 tok/s 40–106 · low confidence	Radeon Instinct MI325X AMD	256 GB	6,000 GB/s	Oct 2024	32.8 GB	Q8_0	Comfortable
58.6 tok/s 35–94 · low confidence	Radeon Instinct MI300A AMD	128 GB	5,325 GB/s	Dec 2023	32.8 GB	Q8_0	Comfortable
58.6 tok/s 35–94 · low confidence	Radeon Instinct MI300X AMD	192 GB	5,325 GB/s	Dec 2023	32.8 GB	Q8_0	Comfortable
58.6 tok/s 35–94 · low confidence	Radeon Instinct MI308X AMD	192 GB	5,325 GB/s	Dec 2023	32.8 GB	Q8_0	Comfortable
55.6 tok/s 33–89 · low confidence	H100 NVL 94 GB NVIDIA	94 GB	3,940 GB/s	Mar 2023	32.8 GB	Q8_0	Comfortable
47.4 tok/s 28–76 · low confidence	H100 PCIe 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	32.8 GB	Q8_0	Comfortable
47.4 tok/s 28–76 · low confidence	H100 SXM5 80 GB NVIDIA	80 GB	3,360 GB/s	Oct 2022	32.8 GB	Q8_0	Comfortable
47.4 tok/s 28–76 · low confidence	H100 SXM5 94 GB NVIDIA	94 GB	3,360 GB/s	Mar 2023	32.8 GB	Q8_0	Comfortable
47.4 tok/s 28–76 · low confidence	H100 SXM5 96 GB NVIDIA	96 GB	3,360 GB/s	Mar 2023	32.8 GB	Q8_0	Comfortable
47.4 tok/s 28–76 · low confidence	H800 SXM5 NVIDIA	80 GB	3,360 GB/s	Mar 2023	32.8 GB	Q8_0	Comfortable
43.7 tok/s 26–70 · low confidence	GeForce RTX 5090 D V2 NVIDIA	24 GB	1,340 GB/s	Aug 2025	18.8 GB	Q4_K_M	Tight
39.8 tok/s 24–64 · low confidence	A30X NVIDIA	24 GB	1,220 GB/s	Apr 2021	18.8 GB	Q4_K_M	Tight
38.4 tok/s 23–61 · low confidence	DRIVE A100 PROD NVIDIA	32 GB	1,870 GB/s	May 2020	25.8 GB	Q6_K	Tight
38.4 tok/s 23–61 · low confidence	GRID A100A NVIDIA	32 GB	1,870 GB/s	May 2020	25.8 GB	Q6_K	Tight
36.7 tok/s 22–59 · low confidence	GeForce RTX 5090 NVIDIA	32 GB	1,790 GB/s	Jan 2025	25.8 GB	Q6_K	Tight
36.7 tok/s 22–59 · low confidence	GeForce RTX 5090 D NVIDIA	32 GB	1,790 GB/s	Jan 2025	25.8 GB	Q6_K	Tight
36.1 tok/s 22–58 · low confidence	Radeon Instinct MI250 AMD	128 GB	3,280 GB/s	Nov 2021	32.8 GB	Q8_0	Comfortable
36.1 tok/s 22–58 · low confidence	Radeon Instinct MI250X AMD	128 GB	3,280 GB/s	Nov 2021	32.8 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: Alibaba
Organisation type: Industry
Country: China
Published: 16 September 2025
Authors: Liangcai Su, Zhen Zhang, Guangyu Li, Zhuo Chen, Chenxi Wang, Maojia Song, Xinyu Wang, Kuan Li, Jialong Wu, Xuanzhong Chen, Zile Qiao, Zhongwang Zhang, Huifeng Yin, Shihao Cai, Runnan Fang, Zhengwei Tao, Wenbiao Yin, Chenxiong Qian, Yong Jiang, Pengjun Xie, Fei Huang, Jingren 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, System control, Quantitative reasoning, Mathematical reasoning, Code generation, Search
Base model: Qwen3-30B-A3B

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: 30B
Training data: 315,000,000,000 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: 6.5 × 10²³ FLOP
How it was established: Operation counting
Fine-tuning compute: 5.7 × 10²¹ FLOP

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: Alibaba-NLP

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

Sources

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

Reference: Scaling Agents via Continual Pre-training
Last updated: 18 December 2025

The extremes

The ten fastest GPUs that run AgentFounder-30B

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 AgentFounder-30B

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

RTX A4500

Memory needed

17.1 GB

Fastest

113 tok/s

AgentFounder-30B reaches a parameter count of 30B. That lands in the range a serious desktop card can handle once the weights are compressed. The number of cards we track that can run it: 132.

The entry point is RTX A4500, with a memory capacity of 20 GB, running it at a compression of IQ4_XS and producing around 22.2 tokens per second.

Top of the range is B200, generating roughly 113 tokens per second on the strength of a memory bandwidth of 8,000 GB/s.

About this model

AgentFounder-30B was published by Alibaba, in the country recorded as China, during September 2025. The category the publisher falls under is industry.

It works in the domain of Language, and is recorded as performing the task of language modeling/generation, Question answering, System control, Quantitative reasoning, Mathematical reasoning, Code generation, Search.

Rather than being trained from scratch, it is derived from Qwen3-30B-A3B. Most models at this scale are adapted from an existing base rather than built from nothing.

Because its weights were released, nothing about running it depends on a provider staying available — it is yours once downloaded. On Hugging Face it is published under the organisation Alibaba-NLP.

How fast it runs, and why

The median result is around 22.0 tokens per second. Exceeding reading speed outright: 104 of them.

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

Producing it required arithmetic totalling around 6.5 × 10²³ FLOP. That figure measures what producing the model cost, and has no bearing on how fast it answers.

Training consumed a corpus of around 315,000,000,000 tokens of text.

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

Step by step

How to choose a GPU for AgentFounder-30B

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 AgentFounder-30B, needing around 17.1 GB at a compression of IQ4_XS. No amount of processing power compensates for a card that cannot hold it.
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 AgentFounder-30B.
03

Choose how far you will compress it

Compression is what makes a model fit smaller cards, at some cost in accuracy, reaching a compression of IQ4_XS on the smallest card that fits. Setting a minimum quality drops the cards that only manage it by squeezing further than you would want, and holds the comparison at one level.
04

Sort by speed

Ranking by tokens per second follows memory bandwidth rather than core counts, for AgentFounder-30B. It will not match a gaming ordering, because generation is bound by memory bandwidth. The card topping the list is B200, at 113 tok/s.
05

Check the fit verdict before buying

The fit column separates cards that just manage it from those with room to spare, in the case of AgentFounder-30B. Comfortable means you can grow the context later. That difference matters more than a few tokens per second, so buy for comfortable if you expect to.
06

Open the card you have settled on

Each card page repeats this sweep for every model we hold, answering what else the hardware is good for beyond AgentFounder-30B.

Answers

AgentFounder-30B — common questions

AgentFounder-30B— how many parameters does it have?

It has a parameter count of 30B. 30B-A3B. 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.

AgentFounder-30B— who created it?

It was published by Alibaba, based in China, an organisation categorised as industry.

AgentFounder-30B— when was it released?

It was published in September 2025.

AgentFounder-30B— what is it used for?

It works in the domain of Language, and is recorded as handling the task of language modeling/generation, Question answering, System control, Quantitative reasoning, Mathematical reasoning, Code generation, Search. 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.

AgentFounder-30B— where can I download it?

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

AgentFounder-30B— how much compute was used to train it?

Training consumed around 6.5 × 10²³ FLOP. 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.

AgentFounder-30B— can I run it if it does not fit in my GPU?

Only by offloading, which is usually a false economy: the part held in system memory drags the whole thing down. The nearest miss we calculate falls short by 4.4 GB. Every figure here assumes the whole model is resident on the card.

AgentFounder-30B— would two GPUs run it faster?

Capacity adds across cards; throughput does not. The number of cards already holding it on their own: 132. So a second card is rarely the answer here.

AgentFounder-30B— why does the quantisation differ between cards?

A larger card holds a more accurate copy. The number of compression levels used across the cards that run it: 5. Bigger cards get the more accurate version, and the quality floor above pins the comparison to one level.

AgentFounder-30B— how accurate are these speed estimates?

They are calculated from specifications rather than measured, and each carries a range. One example: 68–181 tok/s on B200. The same model and card vary by thirty to fifty per cent depending on the inference software and its version.

AgentFounder-30B— what GPU do I need to run it?

The smallest card in our catalogue that holds it is RTX A4500, with a memory capacity of 20 GB. It runs the model at a compression of IQ4_XS using about 17.1 GB, and produces roughly 22.2 tokens per second. The number of cards able to run it in total: 132.

AgentFounder-30B— how fast is it on a GPU?

It depends on the card. The quickest we calculate is B200, at about 113 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and the number of cards clearing that: 104.

AgentFounder-30B— how much VRAM does it need?

It needs about 17.1 GB at a compression of IQ4_XS, 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.

AgentFounder-30B— can I run it on a GPU holding 24 GB?

Yes. The card GeForce RTX 5090 D V2, holding 24 GB, runs it at a compression of Q4_K_M, using about 18.8 GB and generating roughly 43.7 tokens per second. The fit is tight.

AgentFounder-30B— is it open source?

Its weights are published, so it 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.

Source

Original publication

Record last updated 18 December 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