Run unsloth/Qwen3-VL-2B-Instruct-GGUF locally
unsloth/Qwen3-VL-2B-Instruct-GGUF is a compact instruction-tuned chat model with 1.72 billion parameters, built on the qwen3vl architecture. It is released under the apache-2.0 license and has been downloaded 851,190 times.
To run unsloth/Qwen3-VL-2B-Instruct-GGUF locally at a 4,096-token context, its quantized versions need between 1.74 GB (IQ1_S, lowest quality) and 5.21 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is BF16, needing about 5.21 GB. That means unsloth/Qwen3-VL-2B-Instruct-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 |
|---|---|---|---|---|---|---|---|
| IQ1_S | 2.5 | Very low | 0.5 GB | 0.44 GB | 1.74 GB | 798.6 t/s | Fits in VRAM |
| IQ1_M | 2.61 | Low | 0.52 GB | 0.44 GB | 1.76 GB | 764.3 t/s | Fits in VRAM |
| IQ2_XXS | 2.82 | Low | 0.56 GB | 0.44 GB | 1.8 GB | 709.0 t/s | Fits in VRAM |
| IQ2_M | 3.3 | Low | 0.66 GB | 0.44 GB | 1.9 GB | 606.0 t/s | Fits in VRAM |
| IQ3_XXS | 3.56 | Fair | 0.71 GB | 0.44 GB | 1.95 GB | 561.3 t/s | Fits in VRAM |
| Q2_K | 3.62 | Fair | 0.72 GB | 0.44 GB | 1.96 GB | 552.2 t/s | Fits in VRAM |
| Q2_K_L | 3.62 | Fair | 0.72 GB | 0.44 GB | 1.96 GB | 552.2 t/s | Fits in VRAM |
| Q2_K_XL | 3.71 | Fair | 0.74 GB | 0.44 GB | 1.98 GB | 538.3 t/s | Fits in VRAM |
| F16 | 3.81 | Fair | 0.76 GB | 0.44 GB | 2.0 GB | 524.2 t/s | Fits in VRAM |
| Q3_K_S | 4.03 | Fair | 0.81 GB | 0.44 GB | 2.05 GB | 495.2 t/s | Fits in VRAM |
| Q3_K_M | 4.37 | Good | 0.88 GB | 0.44 GB | 2.11 GB | 457.1 t/s | Fits in VRAM |
| Q3_K_XL | 4.51 | Good | 0.9 GB | 0.44 GB | 2.14 GB | 443.3 t/s | Fits in VRAM |
| IQ4_XS | 4.7 | Good | 0.94 GB | 0.44 GB | 2.18 GB | 425.1 t/s | Fits in VRAM |
| IQ4_NL | 4.9 | Good | 0.98 GB | 0.44 GB | 2.22 GB | 407.3 t/s | Fits in VRAM |
| Q4_0 | 4.91 | Good | 0.98 GB | 0.44 GB | 2.22 GB | 406.4 t/s | Fits in VRAM |
| Q4_K_S | 4.93 | Good | 0.99 GB | 0.44 GB | 2.22 GB | 405.1 t/s | Fits in VRAM |
| Q4_K_M | 5.15 | Very good | 1.03 GB | 0.44 GB | 2.27 GB | 387.8 t/s | Fits in VRAM |
| Q4_K_XL | 5.25 | Very good | 1.05 GB | 0.44 GB | 2.29 GB | 380.2 t/s | Fits in VRAM |
| Q4_1 | 5.31 | Very good | 1.06 GB | 0.44 GB | 2.3 GB | 375.9 t/s | Fits in VRAM |
| Q5_K_S | 5.72 | Very good | 1.15 GB | 0.44 GB | 2.38 GB | 349.0 t/s | Fits in VRAM |
| Q5_K_M | 5.85 | Very good | 1.17 GB | 0.44 GB | 2.41 GB | 341.4 t/s | Fits in VRAM |
| Q5_K_XL | 5.86 | Very good | 1.17 GB | 0.44 GB | 2.41 GB | 340.5 t/s | Fits in VRAM |
| Q6_K | 6.59 | Excellent | 1.32 GB | 0.44 GB | 2.56 GB | 302.9 t/s | Fits in VRAM |
| Q6_K_XL | 7.49 | Excellent | 1.5 GB | 0.44 GB | 2.74 GB | 266.6 t/s | Fits in VRAM |
| F32 | 7.57 | Excellent | 1.52 GB | 0.44 GB | 2.75 GB | 263.8 t/s | Fits in VRAM |
| Q8_0 | 8.53 | Excellent | 1.71 GB | 0.44 GB | 2.95 GB | 234.1 t/s | Fits in VRAM |
| Q8_K_XL | 10.85 | Excellent | 2.17 GB | 0.44 GB | 3.41 GB | 184.1 t/s | Fits in VRAM |
| BF16 | 19.85 | Excellent | 3.98 GB | 0.44 GB | 5.21 GB | 100.6 t/s | Fits in VRAM |
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-VL-2B-Instruct-GGUF?
You need about 5.21 GB of VRAM to run unsloth/Qwen3-VL-2B-Instruct-GGUF entirely on the GPU using the BF16 quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run unsloth/Qwen3-VL-2B-Instruct-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run unsloth/Qwen3-VL-2B-Instruct-GGUF fully on the GPU using BF16 (about 5.21 GB).
Can I run unsloth/Qwen3-VL-2B-Instruct-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run unsloth/Qwen3-VL-2B-Instruct-GGUF fully on the GPU using BF16 (about 5.21 GB).
Can I run unsloth/Qwen3-VL-2B-Instruct-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run unsloth/Qwen3-VL-2B-Instruct-GGUF fully on the GPU using BF16 (about 5.21 GB).
What is the best quantization for unsloth/Qwen3-VL-2B-Instruct-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.