RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

Can I run Qwen3.5 27B locally?

Qwen3.5 27B by Qwen has 28B parameters. At 4-bit with an 8K context it needs about 17.0 GB of GPU memory. It fits on 56 of 65 GPUs and Macs we track. See all models that fit in 24 GB of VRAM.

17.0 GBQ4_K_M · 8K context
RX 7900 XTsmallest consumer GPU that fits
M4 32GBsmallest Mac that fits
256Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Qwen3.5-27B-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K)rough10.8 GB11.1 GB12.6 GB18.6 GB
3-bit (Q3_K_M) •usable13.7 GB14.0 GB15.5 GB21.5 GB
4-bit (Q4_K_M) •good, most popular16.7 GB17.0 GB18.5 GB24.5 GB
5-bit (Q5_K_M) •very good19.4 GB19.6 GB21.1 GB27.1 GB
6-bit (Q6_K) •near lossless22.0 GB22.3 GB23.8 GB29.8 GB
8-bit (Q8_0) •effectively lossless27.8 GB28.0 GB29.5 GB35.5 GB
16-bit (F16/BF16) •original weights51.2 GB51.5 GB53.0 GB59.0 GB

Which GPUs and Macs run Qwen3.5 27B?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBTightQ2_K11.1 GB35details
RTX 4060 Ti 16GB16 GBFitsQ2_K11.1 GB28details
RTX 309024 GBFitsQ5_K_M19.6 GB48details
RTX 409024 GBFitsQ5_K_M19.6 GB51details
RTX 509032 GBFitsQ6_K22.3 GB80details
RTX 5070 Ti16 GBFitsQ2_K11.1 GB86details
RX 7900 XTX24 GBFitsQ5_K_M19.6 GB49details
Mac M4 Pro 48GB48 GBFitsQ8_028.0 GB9.5details
Mac M4 Max 128GB128 GBFitsF1651.5 GB10details
Mac M3 Ultra 512GB512 GBFitsF1651.5 GB15details
DGX Spark128 GBFitsF1651.5 GB5.1details
RTX 407012 GBTightQ2_K11.1 GB48details
RTX 5060 Ti 16GB16 GBFitsQ2_K11.1 GB43details
RTX 40608 GBNo———details
Arc B58012 GBTightQ2_K11.1 GB44details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBNo———details
RTX 3080 12GB12 GBTightQ2_K11.1 GB88details
RTX 3090 Ti24 GBFitsQ5_K_M19.6 GB51details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBTightQ2_K11.1 GB48details
RTX 4070 Ti12 GBTightQ2_K11.1 GB48details
RTX 4070 Ti Super16 GBFitsQ2_K11.1 GB65details
RTX 408016 GBFitsQ2_K11.1 GB69details
RTX 4080 Super16 GBFitsQ2_K11.1 GB71details
RTX 50608 GBNo———details
RTX 507012 GBTightQ2_K11.1 GB65details
RTX 508016 GBFitsQ2_K11.1 GB92details
2x RTX 30902×24 GBFitsQ8_028.0 GB33details
2x RTX 40902×24 GBFitsQ8_028.0 GB35details
2x RTX 50902×32 GBFitsF1651.5 GB33details
RTX A600048 GBFitsQ8_028.0 GB27details
RTX 6000 Ada48 GBFitsQ8_028.0 GB34details
RTX PRO 6000 Blackwell96 GBFitsF1651.5 GB33details
L40S48 GBFitsQ8_028.0 GB30details
A100 40GB40 GBFitsQ8_028.0 GB54details
A100 80GB80 GBFitsF1651.5 GB38details
H100 PCIe80 GBFitsF1651.5 GB37details
H100 SXM80 GBFitsF1651.5 GB62details
H200141 GBFitsF1651.5 GB89details
RX 76008 GBNo———details
RX 7700 XT12 GBTightQ2_K11.1 GB41details
RX 7800 XT16 GBFitsQ2_K11.1 GB60details
RX 7900 GRE16 GBFitsQ2_K11.1 GB55details
RX 7900 XT20 GBFitsQ4_K_M17.0 GB48details
RX 907016 GBFitsQ2_K11.1 GB61details
RX 9070 XT16 GBFitsQ2_K11.1 GB61details
Ryzen AI Max+ 395 128GB128 GBFitsF1651.5 GB4.8details
Arc A770 16GB16 GBFitsQ2_K11.1 GB54details
Mac M1 16GB16 GBNo———details
Mac M2 24GB24 GBFitsQ3_K_M14.0 GB7.4details
Mac M3 24GB24 GBFitsQ3_K_M14.0 GB7.4details
Mac M4 16GB16 GBNo———details
Mac M4 32GB32 GBFitsQ4_K_M17.0 GB7.2details
Mac M1 Pro 32GB32 GBFitsQ4_K_M17.0 GB12details
Mac M3 Pro 36GB36 GBFitsQ5_K_M19.6 GB7.6details
Mac M4 Pro 24GB24 GBFitsQ3_K_M14.0 GB20details
Mac M4 Pro 64GB64 GBFitsQ8_028.0 GB9.5details
Mac M1 Max 64GB64 GBFitsQ8_028.0 GB14details
Mac M2 Max 96GB96 GBFitsF1651.5 GB7.4details
Mac M3 Max 128GB128 GBFitsF1651.5 GB7.4details
Mac M4 Max 64GB64 GBFitsQ8_028.0 GB19details
Mac M2 Ultra 192GB192 GBFitsF1651.5 GB15details
Mac M3 Ultra 256GB256 GBFitsF1651.5 GB15details

Prefer an API?

Qwen: Qwen3.5-27B is on OpenRouter at $0.26 per 1M input tokens and $2.60 per 1M output tokens. Price details and history →

Questions

How much VRAM does Qwen3.5 27B need?

About 17.0 GB at 4-bit (Q4_K_M) with an 8K context: 15.6 GB of weights (real GGUF file size), 0.5 GB of KV cache and 0.9 GB of runtime overhead. At 8-bit it needs about 28.0 GB.

What is the smallest GPU that runs Qwen3.5 27B?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RX 7900 XT (20 GB).

Can I run Qwen3.5 27B on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M4 32GB or larger, using the default macOS GPU memory limit.

Similar models

Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.

Qwen3.8 27B

28B · 16.8 GB at 4-bit

Qwen3.6 27B

28B · 17.0 GB at 4-bit

Qwen3 Coder 30B A3B Instruct

31B · 18.7 GB at 4-bit

gemma 3 27b it

27B · 17.3 GB at 4-bit

Arete 27B

27B · 17.4 GB at 4-bit

MiniMax H3 Image PE

27B · 16.8 GB at 4-bit

Model facts

Hugging Face repoQwen/Qwen3.5-27B
Parameters27,781,427,952 (28B)
Layers / hidden size64 / 5120
AttentionGQA (hybrid: 48 linear/state layers); 24 heads, 4 KV heads, head dim 256
KV cache per token (FP16)64.0 KiB
Max context262,144 tokens
Licenseapache-2.0
Released on HFFeb 24, 2026
Downloads (30d) / likes1,678,738 / 1,093

Sources: Hugging Face model API · config.json (Qwen/Qwen3.5-27B) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.