llmrun
AlibabaQwen 3.8Qwen3_5ForConditionalGeneration

Qwen3.8 27B — Hardware Requirements & GPU Compatibility

VisionFunctions

Alibaba's Qwen3.8 27B is a dense, vision-capable model with around 28 billion parameters, continuing the Qwen line into its 3.8 release. It can process images alongside text prompts, supporting tasks like visual question answering and image-grounded chat in addition to general-purpose reasoning and coding assistance. At this size, the model fits on a single high-end consumer GPU once quantized, putting it within reach of enthusiast local setups rather than requiring server-class hardware. It ships with a 256K token context window, enough for long-form documents or extended multi-turn sessions, and is distributed under the Apache 2.0 license. Released in August 2026, it represents one of the more recent entries in Alibaba's Qwen 3.8 family of open-weight models.

27.8B params · dense262K contextApache 2.07.1M downloads17.6K likes24.2M quant downloads
A

Runs well on a 24 GB GPU

Runs from12 GB IQ3_XXS · 16KBest on24 GB Q4_K_M · ~38 tok/s · 64KMacs from32 GB ~5 tok/s

Score 121.5#10 of 92 open#1 of 34 that fit 24 GBApache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

BOn your MacBook Pro 16" M2 Max (32 GB): Q4_K_M fits · 10.6 GB free · ~16 tok/s · up to 128K contextMacBook Pro 16" M2 Max (32 GB) page →

llmrun Composite Score →

121.5

Benchmarks · tier chart ↓
Among 92 open modelsHow the Composite Score is built

Run Qwen3.8 27B

Best fit for your MacBook Pro 16" M2 Max (32 GB). Every other method, OS and quant is in the step-by-step guide.

Recommended for your MacBook Pro 16" M2 Max (32 GB)

MLX
one command · needs mlx-lm installed
$mlx_lm.chat --model lmstudio-community/Qwen3.8-27B-MLX-4bit

MLX repo with its own quantisation

Download: Qwen3.8-27B-UD-Q4_K_M.gguf16.46 GBunsloth

install, OS, other quants

Other ways

each opens step-by-step

Run in cloud

Fits on RTX A5000 (24 GB) (6 GB headroom) · Q4_K_M

Generation speed
~29 tok/s
generation speed
Cost per 1M output tokens
$2.34
per 1M output tokens
Deployon Vast.aiBest optionfrom $0.24/hr (opens in new tab)
Compare GPUs →
or

How Much VRAM Does Qwen3.8 27B Need?

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

QuantizationVRAM+32KOn your device
12.6 GB13.9 GB2-bit · K-quant — 2-bit quantization with K-quant improvementsFits: 256K context · ~22 tok/s
12.9 GB14.2 GB3-bit small — 3-bit small quantizationFits: 256K context · ~22 tok/s
14.3 GB15.6 GB3-bit medium — 3-bit medium quantizationFits: 256K context · ~20 tok/s
14.6 GB16 GB4-bit · legacy — 4-bit legacy quantizationFits: 128K context · ~19 tok/s
17.4 GB18.8 GB4-bit medium — 4-bit medium quantization — most popular sweet spotFits: 128K context · ~16 tok/s
20.5 GB21.9 GB5-bit medium — 5-bit medium quantization — good quality/size tradeoffFits: 128K context · ~14 tok/s
23.7 GB25 GB6-bit, very good — 6-bit quantization, very good qualityFits: 64K context · ~12 tok/s
28.5 GB29.9 GB8-bit, near-lossless — 8-bit quantization, near-losslessDoes not fit: needs 29 GB

Selected Q4_K_M

4-bit medium quantization

17.9GBat 16K context

Weights
16.7 GB
KV cache
0.9 GB
Overhead
0.3 GB

Fits MacBook Pro 16" M2 Max (32 GB) · 10.6 GB free · ~16 tok/s

≥50 tok/s needs ≥1.34 TB/s · RTX 5090 ~67

Which Hardware Can Run Qwen3.8 27B?

Q4_K_M · 17.4 GB

Qwen3.8 27B (Q4_K_M) requires 17.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 23+ GB is recommended. Using the full 262K context window can add up to 13.9 GB, bringing total usage to 31.3 GB. 8 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 125 devices with unified memory can run Qwen3.8 27B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Mini M5 Pro (24 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M
8 GB · RTX 4060Doesn't fit — smallest quant needs 8.8 GB8 GB tier →
12 GB · RTX 3060 12GBIQ3_XXS · 12 GB · ~20 tok/s · 16K12 GB tier →
16 GB · RTX 5060 Ti 16GBQ3_K_L · 15.4 GB · ~19 tok/s · 16K16 GB tier →
24 GB · RTX 4090referenceQ4_K_M · 17.9 GB · ~38 tok/s · 64K24 GB tier →
Mac 32 GB · Mac Mini M4Q4_K_M · 17.9 GB · ~5 tok/s · 128KMacs →
Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 17.9 GB · ~22 tok/s · 256KMacs →

How good is Qwen3.8 27B?

The llmrun Score merges 12 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the 24 GB tier

  • Qwen3.8 27B
  • Other models
  • Best Score for its VRAM

#1 of 34 that fit 24 GBEvery model here fits your MacBook Pro 16" M2 Max (32 GB) (32 GB)

Each dot is a model at the best quant that fits a 24 GB GPU. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.

Full 24 GB ranking →

Benchmarks

The 6 boards that weigh most in the llmrun Score, of 14 with a result for this model; badges show who ran the test.

All 14 Qwen3.8 27B benchmarks

In the 24 GB class

34 models fit 24 GB

Same Score, only models that fit 24 GB at their best quant: the ranking a 24 GB owner actually chooses from.

  1. #1

    Qwen3.8 27BAlibaba

    121.5

    ~38 tok/s · 17.9 GB

  2. #2

    Qwen3.6 27BAlibaba

    111.4

    ~38 tok/s · 17.9 GB

  3. #3

    Qwen3.6 35B A3BAlibaba

    106.9

    ~151 tok/s · 22.0 GB

  4. #4

    Gemma 4 31B ITGoogle

    106.1

    ~32 tok/s · 21.7 GB

  5. #5

    Qwen3.5 35B A3BAlibaba

    105.6

    ~151 tok/s · 22.0 GB

  6. #6

    Gemma 4 26B A4B ITGoogle

    105.4

    ~171 tok/s · 16.5 GB

Full 24 GB ranking →RTX 4090 page →

Alternatives to Qwen3.8 27B

Qwen 3.8 family · 12 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Similar models

10
ModelScoreVRAMtok/sWhy it is similar
Qwen3.6 35B A3BAlibaba106.922 GB~151similar score
Qwen3.5 35B A3BAlibaba105.622 GB~151same size
Gemma 4 26B A4B ITGoogle105.416.5 GB~171same size
Qwen3.5 27BAlibaba103.817.9 GB~38same size
Qwen3 30B A3B Thinking 2507Alibaba100.619.4 GB~173same size
Qwen3 30B A3B Instruct 2507Alibaba96.519.4 GB~173same size
Qwen3 30B A3BAlibaba93.819.4 GB~173same size
Magistral Small 2506Mistral AI88.717.8 GB~44same size
Gemma 3 27B ITGoogle79.321.7 GB~36same size
Muse Glimmer 30BMeta—19.6 GB~36same size
Fine-tunes & variants(38)

Specifications

Publisher
Alibaba
Family
Qwen 3.8
Parameters
27.8B
Architecture
Qwen3_5ForConditionalGeneration
Context Length
262,144 tokens
Vocabulary Size
248,320
Release Date
2026-08-05
License
Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Qwen3.8 27B need?

Qwen3.8 27B requires 17.4 GB of VRAM at Q4_K_M, or 56.3 GB at BF16. Full 262K context adds up to 13.9 GB (31.3 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 27.8B × 4.8 bits ÷ 8 = 16.7 GB

KV Cache + Overhead ≈ 0.7 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 ≈ 14.6 GB (at full 262K context)

VRAM usage by quantization

17.4 GB
31.3 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Qwen3.8 27B?

Yes, at Q6_K (23.7 GB) or lower. Higher quantizations like Q8_0 (28.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Qwen3.8 27B?

For Qwen3.8 27B, Q4_K_M (17.4 GB) offers the best balance of quality and VRAM usage. Q4_K_L (17.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 8.4 GB.

VRAM requirement by quantization

IQ2_XXS
8.4 GB
IQ3_S
12.6 GB
Q3_K_L
15.0 GB
Q4_K_M ★
17.4 GB
Q4_K_L
17.8 GB
BF16
56.3 GB

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Qwen3.8 27B on a Mac?

Yes, but only at lower quantizations. The smallest Mac that can run Qwen3.8 27B is iPad Pro M5 13" (16 GB) at IQ2_XXS; 136 of the 156 Macs we list can run it at some quantization. For Q4_K_M (17.4 GB) you need a Mac with more unified memory.

Can I run Qwen3.8 27B locally?

Yes — Qwen3.8 27B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Qwen3.8 27B?

At Q4_K_M, Qwen3.8 27B can reach ~276 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~38 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = (bandwidth GB/s ÷ model GB) × efficiency

Example: NVIDIA B200 → 8000 ÷ 17.4 × 0.65 = ~299 tok/s

Estimated speed at Q4_K_M (17.4 GB)

~299 tok/s
~38 tok/s
~299 tok/s
~276 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 27B?

At Q4_K_M, the download is about 16.67 GB. The full-precision BF16 version is 55.56 GB. The smallest option (IQ2_XXS) is 7.64 GB.

Which GPUs can run Qwen3.8 27B?

8 consumer GPUs can run Qwen3.8 27B at Q4_K_M. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX. 1 GPU have plenty of headroom for comfortable inference.

Which devices can run Qwen3.8 27B?

125 devices with unified memory can run Qwen3.8 27B at Q4_K_M (17.4 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (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.