Run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF locally
DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF is a large code-focused 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 354,747 times.
To run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF locally at a 4,096-token context, its quantized versions need between 2.14 GB (F16, lowest quality) and 29.09 GB (Q8_0, 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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.
All quantizations
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| F16 | 0.28 | Very low | 0.86 GB | 0.47 GB | 2.14 GB | 463.0 t/s | Fits in VRAM |
| BF16 | 0.28 | Very low | 0.87 GB | 0.47 GB | 2.14 GB | 461.3 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_M | 3.12 | Low | 9.77 GB | 0.47 GB | 11.04 GB | 5.1 t/s | Offload |
| IQ3_M | 3.83 | Fair | 12.01 GB | 0.47 GB | 13.28 GB | 4.2 t/s | Offload |
| IQ4_XS | 4.58 | Good | 14.34 GB | 0.47 GB | 15.61 GB | 3.5 t/s | Offload |
| Q4_K_S | 4.73 | Good | 14.81 GB | 0.47 GB | 16.08 GB | 3.4 t/s | Offload |
| IQ4_NL | 4.79 | Good | 15.01 GB | 0.47 GB | 16.28 GB | 3.3 t/s | Offload |
| Q4_K_M | 5.02 | Very good | 15.7 GB | 0.47 GB | 16.98 GB | 3.2 t/s | Offload |
| Q5_K_S | 5.65 | Very good | 17.69 GB | 0.47 GB | 18.96 GB | 2.8 t/s | Offload |
| Q5_K_M | 5.81 | Very good | 18.2 GB | 0.47 GB | 19.47 GB | 2.7 t/s | Offload |
| Q6_K | 6.66 | Excellent | 20.86 GB | 0.47 GB | 22.13 GB | 2.4 t/s | Offload |
| Q8_0 | 8.88 | Excellent | 27.82 GB | 0.47 GB | 29.09 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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF?
You need about 2.99 GB of VRAM to run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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 DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF fully on the GPU using F32 (about 2.99 GB).
Can I run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF fully on the GPU using IQ4_XS (about 15.61 GB).
Can I run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-GGUF fully on the GPU using Q6_K (about 22.13 GB).
What is the best quantization for DavidAU/Qwen3.6-27B-Heretic-Uncensored-FINETUNE-NEO-CODE-Di-IMatrix-MAX-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.