Run unsloth/Qwen3.6-35B-A3B-MTP-GGUF locally
unsloth/Qwen3.6-35B-A3B-MTP-GGUF is a very large language model with 35.51 billion parameters, built on the qwen35moe architecture. It is released under the apache-2.0 license and has been downloaded 777,715 times.
To run unsloth/Qwen3.6-35B-A3B-MTP-GGUF locally at a 4,096-token context, its quantized versions need between 2.11 GB (F16, lowest quality) and 68.3 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 2.94 GB. That means unsloth/Qwen3.6-35B-A3B-MTP-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 | 0.2 | Very low | 0.84 GB | 0.47 GB | 2.11 GB | 477.6 t/s | Fits in VRAM |
| F32 | 0.4 | Very low | 1.66 GB | 0.47 GB | 2.94 GB | 240.4 t/s | Fits in VRAM |
| IQ1_M | 2.56 | Very low | 10.59 GB | 0.47 GB | 11.86 GB | 4.7 t/s | Offload |
| IQ2_XXS | 2.66 | Low | 11.01 GB | 0.47 GB | 12.28 GB | 4.5 t/s | Offload |
| IQ2_M | 2.68 | Low | 11.07 GB | 0.47 GB | 12.34 GB | 4.5 t/s | Offload |
| Q2_K_XL | 2.83 | Low | 11.71 GB | 0.47 GB | 12.99 GB | 4.3 t/s | Offload |
| IQ3_XXS | 3.17 | Low | 13.1 GB | 0.47 GB | 14.38 GB | 3.8 t/s | Offload |
| IQ3_S | 3.46 | Fair | 14.29 GB | 0.47 GB | 15.57 GB | 3.5 t/s | Offload |
| Q3_K_M | 3.85 | Fair | 15.93 GB | 0.47 GB | 17.2 GB | 3.1 t/s | Offload |
| Q3_K_XL | 3.88 | Fair | 16.04 GB | 0.47 GB | 17.32 GB | 3.1 t/s | Offload |
| IQ4_XS | 4.1 | Fair | 16.96 GB | 0.47 GB | 18.23 GB | 2.9 t/s | Offload |
| IQ4_NL | 4.18 | Fair | 17.26 GB | 0.47 GB | 18.54 GB | 2.9 t/s | Offload |
| Q4_K_S | 4.82 | Good | 19.92 GB | 0.47 GB | 21.19 GB | 2.5 t/s | Offload |
| GGUF | 5.0 | Good | 20.66 GB | 0.47 GB | 21.93 GB | 2.4 t/s | Offload |
| Q4_K_M | 5.11 | Very good | 21.11 GB | 0.47 GB | 22.38 GB | 2.4 t/s | Offload |
| Q4_K_XL | 5.15 | Very good | 21.28 GB | 0.47 GB | 22.56 GB | 2.3 t/s | Offload |
| Q5_K_S | 5.75 | Very good | 23.78 GB | 0.47 GB | 25.06 GB | — | Insufficient |
| Q5_K_M | 6.1 | Very good | 25.23 GB | 0.47 GB | 26.5 GB | — | Insufficient |
| Q5_K_XL | 6.12 | Very good | 25.29 GB | 0.47 GB | 26.57 GB | — | Insufficient |
| Q6_K | 6.76 | Excellent | 27.95 GB | 0.47 GB | 29.22 GB | — | Insufficient |
| Q6_K_XL | 7.35 | Excellent | 30.37 GB | 0.47 GB | 31.65 GB | — | Insufficient |
| Q8_0 | 8.52 | Excellent | 35.21 GB | 0.47 GB | 36.48 GB | — | Insufficient |
| Q8_K_XL | 8.81 | Excellent | 36.41 GB | 0.47 GB | 37.69 GB | — | Insufficient |
| BF16 | 16.22 | Excellent | 67.03 GB | 0.47 GB | 68.3 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/Qwen3.6-35B-A3B-MTP-GGUF?
You need about 2.94 GB of VRAM to run unsloth/Qwen3.6-35B-A3B-MTP-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/Qwen3.6-35B-A3B-MTP-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run unsloth/Qwen3.6-35B-A3B-MTP-GGUF fully on the GPU using F32 (about 2.94 GB).
Can I run unsloth/Qwen3.6-35B-A3B-MTP-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run unsloth/Qwen3.6-35B-A3B-MTP-GGUF fully on the GPU using IQ3_S (about 15.57 GB).
Can I run unsloth/Qwen3.6-35B-A3B-MTP-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run unsloth/Qwen3.6-35B-A3B-MTP-GGUF fully on the GPU using Q4_K_XL (about 22.56 GB).
What is the best quantization for unsloth/Qwen3.6-35B-A3B-MTP-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.