Qwen3.8 27B — VRAM, Speed & Local Setup

作者： Jakub Rusinowski · 最后更新： 2026年9月6日

Model library → Qwen3.8 → Qwen3.8 27B

27.78B dense, natively multimodal — text, images and video in. 262,144-token native context, extendable to 1M with YaRN, Apache 2.0. 17.6 GB at Q4_K_M puts it inside a 24 GB card with room for context, which is what makes it the current default recommendation at the single-GPU tier.

Qwen3.8 27B needs about 18 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.

Specifications

Parameters27.78 Billion
Context window262,144
ArchitectureDense Transformer (native multimodal)
ProviderAlibaba Cloud
LicenceApache 2.0
Specified atQ4_K_M
System RAM32 GB
Record updated2026-09-06

Licence

Apache-2.0 — commercial use permitted. Commercial use permitted. No usage restrictions beyond attribution.

VRAM and Speed by Quantization

Modelled on a reference NVIDIA RTX 4090 (24 GB), with no KV cache (this record has no published architecture). Speed figures are ESTIMATES from the memory-bandwidth roofline described on the methodology page, not benchmarks we ran — rows marked measured come from published or reader-submitted runs. VRAM here includes the KV cache, so it reads higher than the headline figure above, which does not.

QuantWeightsVRAM neededEst. speedFit on 24 GB
Q2_K9.1 GB9.9 GB~65 tok/s (est.)Fits comfortably
Q3_K_M11.8 GB12.6 GB~52 tok/s (est.)Fits comfortably
Q4_K_M16.8 GB17.6 GB~39 tok/s (est.)Fits comfortably
Q5_K_M19.7 GB20.5 GB~34 tok/s (est.)Fits comfortably
Q6_K22.8 GB23.6 GB~30 tok/s (est.)Tight fit
Q8_029.5 GB30.3 GB~4 tok/s (est.)Offloads to system RAM (slow)
F1655.6 GB56.4 GB—Won't fit

Want the memory numbers alone, at every quantization level and your own context length? Use the Qwen3.8 27B VRAM calculator.

Buy This HardwareAMD Radeon RX 7900 XT 20GB — 20 GB VRAM · 315 W board powerDeploy in the Cloud NowRTX 4090 on RunPod — from $0.34/hr · rate checked 2026-07

or compare on Vast.ai from $0.35/hr (typical low · varies)

As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.

Recommended GPU

The cheapest catalogued GPU that runs Qwen3.8 27B is the AMD Radeon RX 7900 XT (20 GB).

联盟营销声明: 本页部分链接为联盟推广链接——如果你通过它们购买，LLM Configurator 可能会获得佣金，而你无需支付任何额外费用。作为亚马逊联盟成员（Amazon Associate），LLM Configurator 会从符合条件的购买中获得收益。
AMD Radeon RX 7900 XT 20GB
20 GB VRAM · 315 W board power
2026年价格波动较大——请以当前商品页价格为准。
在亚马逊查看价格

How to Run Qwen3.8 27B

Install Ollama, then run:

ollama run qwen3.8:27b

Weights on Hugging Face: Qwen/Qwen3.8-27B.

Best for: coding, general purpose, multimodal, long context.

Can I Run Qwen3.8 27B on My GPU?

Other Qwen3.8 Sizes

Qwen3.8 27B — Frequently Asked Questions

How much VRAM does Qwen3.8 27B need?
About 18 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Qwen3.8 27B run on an RTX 4090 (24 GB)?
Yes. Qwen3.8 27B needs about 18 GB at Q4_K_M, inside a 24 GB card, at an estimated 39 tokens/sec.
How do I run Qwen3.8 27B locally?
Install Ollama and run `ollama run qwen3.8:27b`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does Qwen3.8 come in?
Qwen3.8 27B (18 GB), Qwen3.8-Max (1450 GB), Qwen3.8-Flash-Next (109 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.

← All Qwen3.8 models | VRAM calculator | Build a PC for this model | Check your own hardware