llmrun
YoussofalQwen 3.8Qwen4ExpForConditionalGeneration

Qwen3.8 Flash Next MTPLX Optimized Speed — Hardware Requirements & GPU Compatibility

Chat

Qwen3.8 Flash Next MTPLX Optimized Speed is a 126.2B-parameter open language model from Youssofal in the Qwen 3.8 family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 76.10 GB of VRAM — see which GPUs and Macs can run it below.

126.2B params · MoE7.8B active262K contextOther15.3K downloads30 likes
Based on Qwen3.8 Flash Next

Qwen3.8 Flash Next MTPLX Optimized Speed runs on a 96 GB+ Mac or a workstation GPU

Best local pick → Mac Studio (M3 Ultra, 96GB) · Q4_K_M · 76.7 GB · ~56 tok/s

Macs from64 GB Q2_K · ~64 tok/s

Other

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Qwen3.8 Flash Next MTPLX Optimized Speed

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Recommended

vLLM
one command · needs vLLM installed
$vllm serve Youssofal/Qwen3.8-Flash-Next-MTPLX-Optimized-Speed --max-model-len 16384

vLLM repo with its own quantisation

Download: Weights on Hugging Face

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX PRO 6000 (96 GB) (19 GB headroom) · Q4_K_M

Generation speed
~140 tok/s
generation speed
Cost per 1M output tokens
$3.05
per 1M output tokens
Deployon Vast.aiBest optionfrom $1.54/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does Qwen3.8 Flash Next MTPLX Optimized Speed Need?

Select a quantization level; the listed hardware adheres to it.

QuantizationVRAM+32KOn RTX 4090 (reference)
54.0 GB55.3 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsDoes not fit: needs 54.6 GB
61.9 GB63.2 GB3-bit medium — 3-bit medium quantizationDoes not fit: needs 62.5 GB
76.1 GB77.4 GB4-bit medium — 4-bit medium quantization — most popular sweet spotDoes not fit: needs 76.7 GB
90.3 GB91.6 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffDoes not fit: needs 90.9 GB
104.5 GB105.8 GB6-bit, very good — 6-bit quantization, very good qualityDoes not fit: needs 105.1 GB
126.6 GB127.8 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 127.2 GB
252.8 GB254 GB16-bit · bfloat — Brain floating point 16 — preferred for trainingDoes not fit: needs 253.4 GB
est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

76.7GBat 16K context

Weights
75.7 GB
KV cache
0.7 GB
Overhead
0.3 GB

Doesn't fit RTX 4090 (reference) · needs 52.7 GB more

≥50 tok/s needs ≥414 GB/s · M3 Ultra ~56, M2 Ultra ~56

Or rent RTX PRO 6000 from $1.54/hr.

Which Hardware Can Run Qwen3.8 Flash Next MTPLX Optimized Speed?

Q4_K_M · 76.1 GB

Qwen3.8 Flash Next MTPLX Optimized Speed (Q4_K_M) requires 76.1 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 99+ GB is recommended. Using the full 262K context window can add up to 10.7 GB, bringing total usage to 86.8 GB. No consumer GPU has enough memory. 30 devices with unified memory can run Qwen3.8 Flash Next MTPLX Optimized Speed, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8–16 GB · RTX 4060, RTX 3060 12GB, RTX 5060 Ti 16GBDoesn't fit — smallest quant needs 54.6 GB8 GB tier →12 GB tier →16 GB tier →
24 GB · RTX 4090referenceDoesn't fit — smallest quant needs 54.6 GB24 GB tier →
Mac 32 GB · Mac Mini M4Doesn't fit — smallest quant needs 54.6 GBMacs →
Mac 64 GB · Mac Studio M4 MaxQ2_K · 54.6 GB · ~55 tok/s · 128KMacs →

Alternatives to Qwen3.8 Flash Next MTPLX Optimized Speed

Qwen 3.8 family · 12 models →

Similar models

6
ModelScoreVRAMtok/sWhy it is similar
Qwen3.5 122B A10BAlibaba109.475.5 GB—needs > 24 GBsame size
NVIDIA Nemotron 3.5 Super VL 120B A12B BF16NVIDIA—87.7 GB—needs > 24 GBsame size
Ling 3.0 Flash VLInclusion AI—76 GB—needs > 24 GBsame size
Leanstral 1.5 119B A6BMistral AI—84 GB—needs > 24 GBsame size
Laguna S 2.1poolside—72.5 GB—needs > 24 GBsame size
XORTRON.CriminalComputing.LARGE.2026.3darkc0de—79.8 GB—needs > 24 GBsame size

Specifications

Publisher
Youssofal
Family
Qwen 3.8
Parameters
126.2B
Active parameters
7.8B per token
Architecture
Qwen4ExpForConditionalGeneration
Context Length
262,144 tokens
Vocabulary Size
248,320
Release Date
2026-08-28
License
Other

Resources

Frequently Asked Questions

How much VRAM does Qwen3.8 Flash Next MTPLX Optimized Speed need?

Qwen3.8 Flash Next MTPLX Optimized Speed requires 76.1 GB of VRAM at Q4_K_M, or 252.8 GB at BF16. Full 262K context adds up to 10.7 GB (86.8 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 126.2B × 4.8 bits ÷ 8 = 75.7 GB

KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 11 GB (at full 262K context)

VRAM usage by quantization

76.1 GB
86.8 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run Qwen3.8 Flash Next MTPLX Optimized Speed?

No — Qwen3.8 Flash Next MTPLX Optimized Speed requires at least 54.0 GB at Q2_K, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for Qwen3.8 Flash Next MTPLX Optimized Speed?

For Qwen3.8 Flash Next MTPLX Optimized Speed, Q4_K_M (76.1 GB) offers the best balance of quality and VRAM usage. Q5_K_M (90.3 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 54.0 GB.

VRAM requirement by quantization

Q2_K
54.0 GB
Q4_K_M ★
76.1 GB
Q5_K_M
90.3 GB
Q6_K
104.5 GB
Q8_0
126.6 GB
BF16
252.8 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Qwen3.8 Flash Next MTPLX Optimized Speed on a Mac?

Yes — Mac Pro M2 Ultra (64 GB) and 41 other Macs can run Qwen3.8 Flash Next MTPLX Optimized Speed, the smallest at Q2_K. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 76.1 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).

Can I run Qwen3.8 Flash Next MTPLX Optimized Speed locally?

Yes — Qwen3.8 Flash Next MTPLX Optimized Speed can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Qwen3.8 Flash Next MTPLX Optimized Speed?

At Q4_K_M, Qwen3.8 Flash Next MTPLX Optimized Speed can reach ~95 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 4.7 GB active ÷ (8000 × 0.65) = 0.899 ms, plus 48 layers × 0.055 ms = 2.640 ms, so 1000 ÷ 3.539 ms = ~283 tok/s

Estimated speed at Q4_K_M (76.1 GB)

~283 tok/s
~283 tok/s
~243 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Qwen3.8 Flash Next MTPLX Optimized Speed?

At Q4_K_M, the download is about 75.72 GB. The full-precision BF16 version is 252.39 GB. The smallest option (Q2_K) is 53.63 GB.

Which GPUs can run Qwen3.8 Flash Next MTPLX Optimized Speed?

No single consumer GPU has enough VRAM to run Qwen3.8 Flash Next MTPLX Optimized Speed at Q4_K_M (76.1 GB). Multi-GPU or professional hardware is required.

Which devices can run Qwen3.8 Flash Next MTPLX Optimized Speed?

30 devices with unified memory can run Qwen3.8 Flash Next MTPLX Optimized Speed at Q4_K_M (76.1 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.