Qwen3.8-Flash-Next GPU requirements: VRAM and cheapest GPU

Qwen3.8-Flash-Next has 180B parameters, published in BF16. It is a mixture-of-experts model: 10 of its 512 experts work on each token, about 61.6B parameters, so it generates at roughly the speed of a model that size. Every expert still has to sit in GPU memory, which is why the memory figures below are for the full model. Its licence (qwen-community-1.0) adds terms of its own, so read it before you use the model commercially.

Parameters: 180B
Smallest footprint (INT4): about 105 GB
Cheapest GPU per hour (FP16): $2.39
Type: Vision

Qwen/Qwen3.8-Flash-Next

View on Hugging Face (opens in a new tab)Read our review

Parameters 180B
Active per token 61.6B
Type Vision
Task image-text-to-text
Max context 256K
Licence qwen-community-1.0
Downloads 3.1M
Likes 7.1K

More settings

Context length: 8K tokens

Qwen3.8-Flash-Next needs about 397 GB of GPU memory

Inference (vLLM), FP16 / BF16, 8K tokens context, one request at a time.

Where the memory goes

Weights 360 GB
KV cache 0.2 GB
Runtime overhead 37 GB
Total 397 GB

GPU counts keep 10% of each card free for spikes. Plan extra if you expect long prompts or many users at once.

GPUs that fit, cheapest firstEvery GPU type we track that can hold this setup on one node, ranked by total hourly cost
Rank	Configuration	Per GPU / hr	GPUs needed	Total / hr	Total / month (24/7)	Provider
1	3 × MI300X 192GBCheapestCDNA 3 · HBM3	$2.39	3sold as 4 GPUs; you need 3	$9.56	$6,979	RunPod (opens in a new tab)4x MI300X
2	5 × RTX PRO 6000 96GBBlackwell · GDDR7	$1.80	5sold as 8 GPUs; you need 5	$14.40	$10,512	Nebius (opens in a new tab)8x RTX PRO 6000 96GB
3	4 × H200 141GBHopper · HBM3e	$3.79	4	$15.16	$11,067	RunPod (opens in a new tab)4x H200 NVL
4	6 × A100 80GBAmpere · HBM2e	$2.09	6sold as 8 GPUs; you need 6	$16.72	$12,206	RunPod (opens in a new tab)8x A100 PCIe
5	2 × B300 288GBBlackwell Ultra · HBM3e	$8.99	2	$17.98	$13,125	RunPod (opens in a new tab)2x B300
6	6 × H100 80GBHopper · HBM3	$2.89	6sold as 8 GPUs; you need 6	$23.12	$16,878	RunPod (opens in a new tab)8x H100 PCIe
7	5 × H100 NVL 94GBHopper · HBM3	$3.19	5sold as 8 GPUs; you need 5	$25.52	$18,630	RunPod (opens in a new tab)8x H100 NVL
8	3 × B200 180GBBlackwell · HBM3e	$7.79	3sold as 4 GPUs; you need 3	$31.16	$22,747	Lambda (opens in a new tab)4x B200 SXM6
9	3 × GB200 186GBBlackwell · HBM3e	$10.50	3sold as 4 GPUs; you need 3	$42.00	$30,660	CoreWeave (opens in a new tab)GB200 NVL72
More than 8: needs several nodes: L40S 48GB, RTX 6000 Ada 48GB, RTX A6000 48GB, A40 48GB, A100 40GB, RTX 5090 32GB, RTX 4090 24GB, RTX 3090 24GB, L4 24GB, A10 24GB, T4 16GB, V100 16GB

Cost per cloudCheapest offer that fits, per provider. Approximate, prices collected Nov 13, 2026.
Provider	Instance	Per GPU / hr	GPUs needed	Total / hr	Total / month (24/7)
RunPod (opens in a new tab)Cheapest	4x MI300Xus	$2.39	3 × MI300X 192GB sold as 4 GPUs; you need 3	about $9.56	about $6,979
Nebius (opens in a new tab)	8x RTX PRO 6000 96GBeu/us	$1.80	5 × RTX PRO 6000 96GB sold as 8 GPUs; you need 5	about $14.40	about $10,512
CoreWeave (opens in a new tab)	RTX PRO 6000 Blackwell Server Edition (High Memory)us	$2.50	5 × RTX PRO 6000 96GB sold as 8 GPUs; you need 5	about $20.00	about $14,600
Vultr (opens in a new tab)	vbm-112c-2048gb-8-a100-gpu (bare metal)us	$2.80	6 × A100 80GB sold as 8 GPUs; you need 6	about $22.40	about $16,352
Vast.ai (opens in a new tab)	4x H200 (median of 4 offers)worldwide marketplace	$5.67	4 × H200 141GB	about $22.68	about $16,554
Lambda (opens in a new tab)	8x HGX A100us	$3.19	6 × A100 80GB sold as 8 GPUs; you need 6	about $25.52	about $18,630
AWS (opens in a new tab)	p4de.24xlargeus-east-1	$3.43	6 × A100 80GB sold as 8 GPUs; you need 6	about $27.45	about $20,036
Azure (opens in a new tab)	ND96amsr_A100_v4eastus	$4.10	6 × A100 80GB sold as 8 GPUs; you need 6	about $32.77	about $23,922
Google Cloud (opens in a new tab)	g4-standard-384us-central1	$4.50	5 × RTX PRO 6000 96GB sold as 8 GPUs; you need 5	about $36.00	about $26,280

Prices collected Nov 13, 2026. Every cost is approximate: on-demand, US dollars, mostly US regions (marketplaces such as Vast.ai are worldwide).

Memory by precision and workload

With a 8K-token context and one request at a time. LoRA and full fine-tuning start from 16-bit weights and QLoRA from 4-bit, so the other cells are blank.
Precision	Inference	LoRA fine-tune	QLoRA fine-tune	Full fine-tune
FP16 / BF16	397 GB	463 GB	–	3,172 GB
INT8	209 GB	–	–	–
INT4	105 GB	–	171 GB	–

Cheapest GPU by precision

Inference with an 8K context and one request at a time.

FP16 / BF16

about 397 GB · 8K tokens

The weights as published. Nothing lost, but the largest footprint.

3 × MI300X 192GB

RunPod (opens in a new tab) · 4x MI300X

sold as 4 GPUs; you need 3

Per GPU / hr: $2.39
GPUs needed: 3
Total / hr: about $9.56
per month, 24/7, approx.: $6,979

INT8

about 209 GB · 8K tokens

Weights at half the size. Most chat and RAG work shows no visible difference.

5 × A40 48GB

RunPod (opens in a new tab) · 8x A40

sold as 8 GPUs; you need 5

Per GPU / hr: $0.59
GPUs needed: 5
Total / hr: about $4.72
per month, 24/7, approx.: $3,446

INT4

about 105 GB · 8K tokens

Weights at a quarter of the size. Small drops show up first in maths and code.

3 × A40 48GB

RunPod (opens in a new tab) · 4x A40

sold as 4 GPUs; you need 3

Per GPU / hr: $0.59
GPUs needed: 3
Total / hr: about $2.36
per month, 24/7, approx.: $1,723

Inference vs fine-tuning

Inference (FP16)
1×
of the memory inference needs
397 GB
3 × MI300X 192GB · about $9.56
LoRA fine-tune
1.2×
of the memory inference needs
463 GB
3 × MI300X 192GB · about $9.56
QLoRA fine-tune
0.4×
of the memory inference needs
171 GB
4 × A40 48GB · about $2.36
Full fine-tune
8×
of the memory inference needs
3,172 GB
more than one 8-GPU node

Serving it at FP16 fits on 3 × MI300X 192GB (RunPod). LoRA needs 3 × MI300X 192GB (RunPod), QLoRA 4 × A40 48GB (RunPod) and a full fine-tune more than one 8-GPU node. Training keeps activations and optimiser state in memory as well as the weights, which is why it so often needs more GPUs than serving the same model.

Will it run on a Mac or a single RTX 4090?

RTX 4090 (24 GB)

Not on its own. Even at INT4 it needs about 105 GB, more than the card’s 24 GB. Two or more cards, or a bigger data-centre GPU, will do it.

Mac with Apple silicon

Yes, with enough memory. At INT4 it needs about 105 GB, so a Mac with 192 GB or more of unified memory (macOS lets the GPU use about three quarters of it). Expect slower generation than on a data-centre GPU.

Run it yourself or pay per token?

Run it yourself

The cheapest setup we found at FP16 is 3 × MI300X 192GB on RunPod, about $6,979 a month if it runs all the time. Worth it when traffic is steady or the data must stay on your own machines.

Use a hosted API

Our review of Qwen3.8-Flash-Next covers where to use it through an API and what that costs. Paying per token usually works out cheaper for light or uneven traffic.

Read our review

Similar models

Open models of a similar size.

Calculate another model

Qwen3.8-Flash-Next: common questions

How much VRAM does Qwen3.8-Flash-Next need?

About 397 GB at FP16, 209 GB at INT8 and 105 GB at INT4, with an 8K context and one request at a time. Longer contexts and more parallel requests add KV cache on top.

What is the cheapest GPU to run Qwen3.8-Flash-Next?

3 × MI300X 192GB (about $2.39 per GPU / hour; sold as 4 GPUs; you need 3) on RunPod, at about $9.56 an hour for FP16 / BF16 inference with an 8K context. Prices collected Nov 13, 2026; they change often.

Can I fine-tune Qwen3.8-Flash-Next on a single GPU?

Not on one GPU. Even QLoRA needs about 171 GB, more than an 80 GB GPU holds. LoRA needs about 463 GB and a full fine-tune about 3,172 GB.

How are these numbers calculated?

Weights are parameters times bytes per parameter (2 at FP16, 1 at INT8, 0.5 at INT4). The KV cache comes from the model’s layers, KV heads, head size and context length. We add about 10% plus 1 GB for the runtime and fill each GPU to 90%. Prices are on-demand list prices in US regions, collected from each provider’s public pricing.

Can Qwen3.8-Flash-Next run on a consumer GPU like the RTX 4090?

Not on its own. Even at INT4 it needs about 105 GB, more than the card’s 24 GB. Two or more cards, or a bigger data-centre GPU, will do it.

Does Qwen3.8-Flash-Next fit on a single H100 80GB?

No. Even at INT4 it needs about 105 GB, so you need several GPUs.

How many GPUs do I need for Qwen3.8-Flash-Next?

At FP16 with an 8K context: 6 × 80 GB, 10 × 48 GB or 19 × 24 GB GPUs. At INT4 you need about a quarter of the memory for the weights.

How much VRAM does Qwen3.8-Flash-Next need at INT8?

About 209 GB: 189 GB for the weights, 0.2 GB for an 8K KV cache and 20 GB of runtime overhead.

How much does it cost to run Qwen3.8-Flash-Next 24/7 for a month?

About $6,979 on the cheapest setup we found (3 × MI300X 192GB (about $2.39 per GPU / hour; sold as 4 GPUs; you need 3), $9.56 an hour for 730 hours). The cheapest on AWS, Google Cloud or Azure is about $20,036 a month.

FP16 vs INT8 vs INT4: which should I use?

FP16 keeps full quality and needs the most memory (about 397 GB here). INT8 halves the weights with little loss for most chat work. INT4 (AWQ, GPTQ or GGUF Q4) cuts them to a quarter, about 105 GB in total, with a small drop in quality that shows most in maths and code.

Can Qwen3.8-Flash-Next run on a Mac?

vLLM or llama.cpp?

vLLM is built for serving many requests at once on NVIDIA and AMD GPUs and reserves most of the GPU memory for KV cache. llama.cpp runs GGUF files on almost anything, including Macs and CPUs, and suits one user or a small team. Memory needs are similar for the same precision.