Advisor / Models / Qwen3.6 27B

Qwen3.6 27B: VRAM requirements & speed

Qwen3.6 27B (27.78B dense, apache-2.0). Native context 256k, 1.01M with YaRN. Smallest GGUF: 16.82 GB.

16.82 GBsmallest quant file
18hardware pages that fit it fully
7run with offload
6measured data points

Memory needed by quantization and context

GGUF file size plus f16 KV cache (and fixed overhead of ~1 GB per GPU not shown). Source: unsloth/Qwen3.6-27B-GGUF.

QuantFile+8k ctx+32k ctx+128k ctx+256k ctx+1M ctx
Q4_K_M16.82 GB17.5 GB19.1 GB25.6 GB34.2 GB82.5 GB
Q5_K_M19.51 GB20.2 GB21.8 GB28.3 GB36.8 GB85.2 GB
Q6_K22.52 GB23.2 GB24.8 GB31.3 GB39.9 GB88.2 GB
Q8_028.6 GB29.3 GB30.9 GB37.3 GB45.9 GB94.3 GB

KV cache size by context length

Computed from the model's attention layout, including a fixed 0.16 GB linear-attention state.

Contextf16q8_0q4_0
8k0.69 GB0.44 GB0.31 GB
32k2.30 GB1.30 GB0.76 GB
128k8.75 GB4.72 GB2.57 GB
256k17.34 GB9.28 GB4.99 GB
1M YaRN65.69 GB34.97 GB18.59 GB

Which hardware runs Qwen3.6 27B? · 16k context

HardwareQuantFitMemoryTok/sSpeed source
RTX 5090 32GBQ8_0fits fully29.8 GB49scaled from measured
RTX 4090 24GBQ5_K_Mfits fully20.7 GB42scaled from measured
RTX PRO 6000 96GB (bonus page)Q8_0fits fully29.8 GB39estimated
RTX 3090 24GBQ5_K_Mfits fully20.7 GB36scaled from measured
AMD RX 7900 XTX 24GBQ5_K_Mfits fully20.7 GB30scaled from measured
Mac Studio M5 Ultra 256GBQ8_0fits fully29.8 GB26estimated
Mac Studio M5 Ultra 512GB (ships late Oct 2026)Q8_0fits fully29.8 GB26estimated
2x RTX 3090 (48GB total)Q8_0fits fully29.8 GB22estimated
Mac Studio M5 Max 128GBQ8_0fits fully29.8 GB17scaled from measured
MacBook Pro 14/16 M5 Max 128GBQ8_0fits fully29.8 GB17scaled from measured
Mac Studio (2025) M3 Ultra 96GBQ5_K_Mfits fully20.7 GB17estimated
MacBook Pro 14/16 (2024) M4 Max 128GBQ4_K_Mfits fully18.1 GB15estimated
RTX 5080 16GBQ4_K_Mpartial offload (slow)18.1 GB14estimated
AMD Ryzen AI Max+ 395 (Strix Halo) 128GBQ4_K_Mfits fully18.1 GB13measured
RTX 5070 Ti 16GBQ4_K_Mpartial offload (slow)18.1 GB13estimated
NVIDIA DGX Spark 128GBQ4_K_Mfits fully18.1 GB13estimated
MacBook Pro 14/16 M5 Pro 48GBQ4_K_Mfits fully18.1 GB13estimated
16GB VRAM class (RTX 4060 Ti 16GB, 32GB RAM)Q4_K_Mpartial offload (slow)18.1 GB10estimated
Mac mini (2024) M4 Pro 64GBQ4_K_Mfits fully18.1 GB9estimated
RTX 5070 12GBQ4_K_Mpartial offload (slow)18.1 GB7estimated
RTX 4070 12GBQ4_K_Mpartial offload (slow)18.1 GB7estimated
Mac mini M6 32GBQ4_K_Mfits fully18.1 GB6estimated
MacBook Air M5 32GBQ4_K_Mfits fully18.1 GB6estimated
8GB VRAM class (RTX 4060 8GB, 32GB RAM)Q4_K_Mpartial offload (slow)18.1 GB4estimated
RTX 3060 12GB (32GB RAM)Q4_K_Mpartial offload (slow)18.1 GB4estimated
Mac mini / iMac M4 16GB-does not fit--

Discrete GPUs are modelled with 32 GB of system RAM available for offload. Does not fit even with offload: Mac mini / iMac M4 16GB.

Links

Model card · GGUF files · Open in the advisor · last verified 2026-10-14

FAQ

How much VRAM does Qwen3.6 27B need?

The smallest quant (Q4_K_M) is 16.82 GB, so you need roughly 19 GB for an 8k context including KV cache and overhead. At 1M context the f16 KV cache alone is 65.7 GB.

Can I run Qwen3.6 27B on a RTX 5090 32GB?

Yes: on RTX 5090 32GB it fits fully as Q8_0 at about 49 tok/s (scaled from measured). 18 of 26 hardware configurations fit it fully at 16k context.

What is the maximum context for Qwen3.6 27B?

256k tokens natively, up to 1.01M with YaRN as documented on the model card. Memory, not the model, is usually the limit: see the KV cache table above.

All models: model index · by memory: 8 GB · 12 GB · 16 GB · 24 GB · 32 GB · 48 GB · 64 GB · 96 GB · 128 GB · 192 GB · 256 GB · 512 GB. Numbers use the same fit math as the advisor; sources and methods.