Qwen / qwen4_exp_text

Qwen3.8-Flash-Next

Name: Qwen3.8-Flash-Next architecture and memory requirements
Creator: LLM Bottleneck
License: https://creativecommons.org/licenses/by/4.0/

180.0 billion parameters, routing 10 of 512 experts per token.

Open in the calculator →

Architecturepublished data

Quick answer

Qwen3.8-Flash-Next needs about 111.6 GB at Q4_K_M — the format most people download — with 8,192 tokens of context. It fits entirely on 5 of 18 common devices, starting with the AMD Ryzen AI Max+ 395 with Radeon 8060S (128 GB). What else fits in 128 GB. No hardware for it? Rent a GPU that runs it, priced live.

Smallest device that fitsAMD Ryzen AI Max+ 395 with Radeon 8060S128 GB · 256 GB/s~55 tok/s Faster than you readOpen in the calculator →
Fastest in this listApple M2 Ultra192 GB · 800 GB/s~95 tok/s Faster than you readOpen in the calculator →

Sized at 8,192 tokens of context with the whole model in device memory, across 18 common devices. Speeds are estimates, not benchmarks.

Find the AMD Ryzen AI Max+ 395 with Radeon 8060S, the smallest card listed here that holds it: Amazon ↗ · eBay (new and used) ↗Store links may pay us a commission. They never decide which card is suggested — the memory arithmetic does.

Weight-file size by format

every format, with its evidence

Format	Size	Range	Basis
FP16	360.2 GB	252.0 GB – 363.6 GB	size range
Q8_0	191.4 GB	133.9 GB – 194.4 GB	size range
Q6_K	147.8 GB	103.4 GB – 150.6 GB	size range
Q5_K_M	128.6 GB	86.6 GB – 150.6 GB	size range
Q5_0	125.6 GB	86.6 GB – 150.6 GB	size range
Q4_K_M	110.5 GB	70.9 GB – 150.6 GB	size range
Q4_0	104.7 GB	70.9 GB – 150.6 GB	size range
Q3_K_M	90.0 GB	54.1 GB – 150.6 GB	size range
Q2_K	71.3 GB	41.3 GB – 150.6 GB	size range

A file size is not the memory a run needs: the KV cache and the runtime reserve come on top, and the calculator adds both.

What the labels mean

published data: Read from a published source — a file's byte count, a model's configuration or a manufacturer's specification — or exact arithmetic on such values. Not a measurement on a machine.
reconstructed size: Weight size reconstructed from the pinned architecture, because no published file exists.
size range: Only a lower and an upper bound are claimed for this weight size.

Hardware ladder

Q4_K_M at 8,192 tokens · 5 of 18 devices hold it entirely

Device	Memory	Verdict	Decode	Calculator
NVIDIA GeForce RTX 5070	111.6 GB of 12 GB	44 layers on system RAM	~19 tok/s with offload	Open →
AMD Radeon™ RX 9070 XT	111.6 GB of 16 GB	42 layers on system RAM	~20 tok/s with offload	Open →
NVIDIA GeForce RTX 4080	111.6 GB of 16 GB	42 layers on system RAM	~20 tok/s with offload	Open →
NVIDIA GeForce RTX 5080	111.6 GB of 16 GB	42 layers on system RAM	~20 tok/s with offload	Open →
NVIDIA GeForce RTX 5070 Ti	111.6 GB of 16 GB	42 layers on system RAM	~20 tok/s with offload	Open →
NVIDIA GeForce RTX 5060 Ti 16GB	111.6 GB of 16 GB	42 layers on system RAM	~20 tok/s with offload	Open →
NVIDIA GeForce RTX 4090	111.6 GB of 24 GB	39 layers on system RAM	~22 tok/s with offload	Open →
AMD Radeon™ RX 7900 XTX	111.6 GB of 24 GB	39 layers on system RAM	~22 tok/s with offload	Open →
NVIDIA GeForce RTX 3090	111.6 GB of 24 GB	39 layers on system RAM	~22 tok/s with offload	Open →
NVIDIA GeForce RTX 5090	111.6 GB of 32 GB	35 layers on system RAM	~25 tok/s with offload	Open →
NVIDIA RTX 6000 Ada Generation	111.6 GB of 48 GB	28 layers on system RAM	~29 tok/s with offload	Open →
Apple M4 Pro	111.6 GB of 64 GB	does not fit	not calibrated	Open →
Apple M5 Pro	111.6 GB of 64 GB	does not fit	not calibrated	Open →
AMD Ryzen AI Max+ 395 with Radeon 8060S	111.6 GB of 128 GB	fits	~55 tok/s	Open →
Apple M3 Max	111.6 GB of 128 GB	fits	~65 tok/s	Open →
Apple M4 Max	111.6 GB of 128 GB	fits	~79 tok/s	Open →
NVIDIA DGX Spark	111.6 GB of 128 GB	fits	~52 tok/s	Open →
Apple M2 Ultra	111.6 GB of 192 GB	fits	~95 tok/s	Open →

Decode is a calibrated estimate from the published calibration; capacity is the manufacturer's published ceiling, not guaranteed free memory. Fitting in memory is not the same as loading: whether the runtime and version you have supports this architecture and format on that machine has not been tested here. Offloaded rows assume there is enough system RAM for the overflow — the calculator checks that against the RAM you declare. Each “Open” link keeps this model, format and context.

Run it in the cloud

Rent a GPU that runs Qwen3.8-Flash-Next

Every rentable machine that holds the whole model, from Vast.ai and RunPod, with this site's speed estimate for each and the provider's own price, read live. Cheapest first; sort by cost per token or speed instead, or switch the format.

Prices · 18:30 UTC, 22 Sept

Machine	Speed	Per hour	Per M tokens	Where	Rent
4× RTX 5090128 GBrecommendedcheapestbest value	~1172 tok/sFaster than you read	$1.07	$0.25	Vast.aiMarketplace · 99.6% reliable	Rent on Vast.ai
8× RTX 5060 Ti 16GB128 GB	~483 tok/sFaster than you read	$1.60	$0.92	Vast.aiMarketplace · 99.7% reliable	Rent on Vast.ai
4× L40S192 GB	~565 tok/sFaster than you read	$1.87	$0.92	Vast.aiMarketplace · 99.2% reliable	Rent on Vast.ai
2× RTX PRO 6000192 GB	~655 tok/sFaster than you read	$2.27	$0.96	Vast.aiMarketplace · 98.8% reliable	Rent on Vast.ai
4× RTX 6000 Ada192 GB	~628 tok/sFaster than you read	$3.00	$1.33	Vast.aiMarketplace · 98.5% reliable	Rent on Vast.ai
4× RTX PRO 5000192 GB	~879 tok/sFaster than you read	$3.28	$1.04	RunPodCommunity Cloud	Rent on RunPod
2× H100 SXM160 GB	~1225 tok/sFaster than you read	$3.47	$0.79	Vast.aiMarketplace · 98.7% reliableRunPod $6.98/h	Rent on Vast.ai
1× H200141 GB	~907 tok/sFaster than you read	$3.59	$1.10	RunPodCommunity Cloud	Rent on RunPod

No GPU at all

Use Qwen3.8-Flash-Next by the token

Novita AI$0.47 / M output · $0.15 / M inputOpen on Novita AI
OpenRouter$0.47 / M output · $0.15 / M inputOpen on OpenRouter

Here the rented card is as cheap as the API or cheaper per token ($0.25 per million), and it keeps your data on a machine you control. The API is still the easier start: nothing to set up and nothing to switch off.

First time renting a GPU? How it works, in four steps

Create an account and add credit. Both providers are prepaid: $10 is enough to try any card on this page for hours. A new RunPod account that arrives through a referral link gets a one-time $5 credit when it first adds $10.
Pick the card and the count shown here, and a template: llama.cpp or vLLM for an OpenAI-compatible API, or one with a web chat if you only want to talk to the model. “How to launch it” below gives the exact image and command.
Wait for the download. The weights are fetched on the machine; tens of gigabytes take a few minutes on a data-centre connection. You pay by the second from the moment it starts.
Stop it when you are done. A running machine bills even when idle. A stopped one costs nothing per hour, though its disk is usually still billed until you delete it.

How to launch it

No Q4_K_M file of Qwen3.8-Flash-Next is catalogued here, so there is no exact command to vouch for. Start the machine from the llama.cpp server image ghcr.io/ggml-org/llama.cpp:server-cuda and point -hf at a Q4_K_M build from the publisher or a quantizer you trust — search Hugging Face for one. Check its file size against the memory figure on this page before you rent.

Every machine holds the whole model at Q4_K_M and 8,192 tokens of context, no offload. Speeds are this site’s single-stream decode estimates; prices are what each provider’s own API quoted, on-demand, for the whole machine. Vast hosts below 98% measured reliability are left out.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Buying a card instead, or paying by the token? Compare a year of Qwen3.8-Flash-Next three ways →

Under the hood

Architecture, read from the publisher’s file

The numbers every figure above is computed from, with the file they came from.

Architecture

architecture.json · pinned de4b8e4d43b9

✓ Architecture read from the published config.json

Retrieved 2026-09-01 at pinned commit de4b8e4d43b9; parameter count from the safetensors index at the same revision.

Architecture: qwen4_exp_text
Layers: 48
Hidden size: 2,560
Attention heads: 24
KV heads: 2
Head dimension: 256
Vocabulary: 248,320
Context ceiling: 262,144
RoPE theta: 10,000,000
Experts: 512
Experts / token: 10
Expert width: 640
Shared expert width: 640

Where the memory goes

Grouped-query attention shares each key/value head across 12 query heads, so the KV cache is 8% of what multi-head attention would need at the same context.

This is a multimodal checkpoint (vision); the diagram and memory sizing above cover the text decoder. Encoder/audio/image activations are not included in the KV or weight figures.

Open the exact official config.json ↗

Qwen3.8-Flash-Next, card by card

One page per device: whether Qwen3.8-Flash-Next fits, at which formats, and how fast.

All 32Fewer

Derivatives this page also answers for

Each of these repositories declares Qwen3.8-Flash-Next as its base, and its published configuration matches this one on every field that decides memory. Its page lists what was compared and any files it publishes. Paste any other repository.

ukisai/Swift1.5-Qwen3.8-Flash-Next badtheorylabs/Tinfield-1 Blackfrost-AI/CYBER-FROST-3.8-BF16 beamster/Qwen3.8-Flash-Next-Sushi-4bpw beamster/Qwen3.8-Flash-Next-Sushi-2.6bpw beamster/Qwen3.8-Flash-Next-Sushi-2bpw

Use this answer in your own product

As JSON, from the API

curl -s https://llmbottleneck.com/v1/analyze \
  -H "Authorization: Bearer $LLMB_KEY" -H "content-type: application/json" \
  -d '{"model":"qwen-qwen3-8-flash-next","quantization":"Q4_K_M","context":8192,"hardware":"nvidia-geforce-rtx-4090"}'

Same engine, same evidence, every field sourced. Free key in one step →

On your page, as a widget

<script src="https://llmbottleneck.com/widget.js"
  data-model="qwen-qwen3-8-flash-next" data-quantization="Q4_K_M"
  data-hardware="nvidia-geforce-rtx-4090" data-context="8192"></script>

No key needed. Unbranded, with your own buy button, on Pro and Business →

Citing this page

LLM Bottleneck. “Qwen3.8-Flash-Next VRAM and hardware requirements.” Architecture from Qwen/Qwen3.8-Flash-Next at revision de4b8e4d43b9, retrieved 2026-09-01. https://llmbottleneck.com/models/qwen-qwen3-8-flash-next

Every figure above is either the published value or a reconstruction whose measured error is on the accuracy page.