Run unsloth/Qwen3.6-27B-GGUF locally

License: apache-2.0 ⬇ 578,691 ❤ 829
Parameters26.9B
Context262,144

unsloth/Qwen3.6-27B-GGUF is a large language model with 26.9 billion parameters, built on the qwen35 architecture. It is released under the apache-2.0 license and has been downloaded 578,691 times.

To run unsloth/Qwen3.6-27B-GGUF locally at a 4,096-token context, its quantized versions need between 2.14 GB (F16, lowest quality) and 52.25 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.

For most users the best balance is F32, needing about 2.99 GB. That means unsloth/Qwen3.6-27B-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.

→ Guide: How much VRAM do you need?

All quantizations

Quant.Bits QualityWeights KVTotal Speed~Verdict
F16 0.28 Very low 0.86 GB 0.47 GB 2.14 GB 463.0 t/s Fits in VRAM
F32 0.55 Very low 1.72 GB 0.47 GB 2.99 GB 233.0 t/s Fits in VRAM
IQ2_XXS 2.79 Low 8.74 GB 0.47 GB 10.02 GB 5.7 t/s Offload
IQ2_M 3.23 Low 10.1 GB 0.47 GB 11.38 GB 4.9 t/s Offload
Q2_K_XL 3.52 Fair 11.04 GB 0.47 GB 12.31 GB 4.5 t/s Offload
IQ3_XXS 3.57 Fair 11.17 GB 0.47 GB 12.45 GB 4.5 t/s Offload
Q3_K_S 3.68 Fair 11.51 GB 0.47 GB 12.78 GB 4.3 t/s Offload
Q3_K_M 4.04 Fair 12.65 GB 0.47 GB 13.93 GB 4.0 t/s Offload
Q3_K_XL 4.31 Good 13.48 GB 0.47 GB 14.75 GB 3.7 t/s Offload
IQ4_XS 4.59 Good 14.38 GB 0.47 GB 15.65 GB 3.5 t/s Offload
Q4_0 4.7 Good 14.71 GB 0.47 GB 15.98 GB 3.4 t/s Offload
Q4_K_S 4.72 Good 14.77 GB 0.47 GB 16.04 GB 3.4 t/s Offload
IQ4_NL 4.78 Good 14.97 GB 0.47 GB 16.24 GB 3.3 t/s Offload
Q4_K_M 5.0 Very good 15.66 GB 0.47 GB 16.94 GB 3.2 t/s Offload
Q4_1 5.13 Very good 16.07 GB 0.47 GB 17.34 GB 3.1 t/s Offload
Q4_K_XL 5.24 Very good 16.4 GB 0.47 GB 17.68 GB 3.0 t/s Offload
Q5_K_S 5.64 Very good 17.66 GB 0.47 GB 18.93 GB 2.8 t/s Offload
Q5_K_M 5.8 Very good 18.17 GB 0.47 GB 19.44 GB 2.8 t/s Offload
Q5_K_XL 5.96 Very good 18.66 GB 0.47 GB 19.94 GB 2.7 t/s Offload
Q6_K 6.7 Excellent 20.98 GB 0.47 GB 22.25 GB 2.4 t/s Offload
Q6_K_XL 7.63 Excellent 23.88 GB 0.47 GB 25.15 GB Insufficient
Q8_0 8.51 Excellent 26.63 GB 0.47 GB 27.91 GB Insufficient
Q8_K_XL 10.51 Excellent 32.9 GB 0.47 GB 34.17 GB Insufficient
BF16 16.28 Excellent 50.98 GB 0.47 GB 52.25 GB Insufficient

KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

How much VRAM do you need to run unsloth/Qwen3.6-27B-GGUF?

You need about 2.99 GB of VRAM to run unsloth/Qwen3.6-27B-GGUF entirely on the GPU using the F32 quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run unsloth/Qwen3.6-27B-GGUF on an 8 GB GPU?

Yes. With 8 GB of VRAM you can run unsloth/Qwen3.6-27B-GGUF fully on the GPU using F32 (about 2.99 GB).

Can I run unsloth/Qwen3.6-27B-GGUF on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run unsloth/Qwen3.6-27B-GGUF fully on the GPU using Q4_0 (about 15.98 GB).

Can I run unsloth/Qwen3.6-27B-GGUF on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run unsloth/Qwen3.6-27B-GGUF fully on the GPU using Q6_K (about 22.25 GB).

What is the best quantization for unsloth/Qwen3.6-27B-GGUF?

If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.