Run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF locally
MaziyarPanahi/Phi-3.5-mini-instruct-GGUF is a mid-size instruction-tuned chat model with 3.82 billion parameters, built on the phi3 architecture. It has been downloaded 240,070 times.
To run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF locally at a 4,096-token context, its quantized versions need between 3.08 GB (IQ1_S, lowest quality) and 6.08 GB (Q8_0, highest quality) of memory, weights plus KV cache and a system margin included.
For most users the best balance is Q3_K_S, needing about 3.87 GB. That means MaziyarPanahi/Phi-3.5-mini-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 | 1.76 | Very low | 0.78 GB | 1.5 GB | 3.08 GB | 510.3 t/s | Fits in VRAM |
| IQ1_M | 1.92 | Very low | 0.85 GB | 1.5 GB | 3.15 GB | 468.3 t/s | Fits in VRAM |
| IQ2_XS | 2.41 | Very low | 1.07 GB | 1.5 GB | 3.37 GB | 372.5 t/s | Fits in VRAM |
| Q2_K | 2.97 | Low | 1.32 GB | 1.5 GB | 3.62 GB | 303.3 t/s | Fits in VRAM |
| IQ3_XS | 3.4 | Fair | 1.51 GB | 1.5 GB | 3.81 GB | 264.3 t/s | Fits in VRAM |
| Q3_K_S | 3.52 | Fair | 1.57 GB | 1.5 GB | 3.87 GB | 255.4 t/s | Fits in VRAM |
| Q3_K_M | 4.09 | Fair | 1.82 GB | 1.5 GB | 4.12 GB | 27.5 t/s | Offload |
| IQ4_XS | 4.31 | Good | 1.92 GB | 1.5 GB | 4.22 GB | 26.1 t/s | Offload |
| Q3_K_L | 4.37 | Good | 1.94 GB | 1.5 GB | 4.24 GB | 25.7 t/s | Offload |
| Q4_K_S | 4.58 | Good | 2.04 GB | 1.5 GB | 4.34 GB | 24.5 t/s | Offload |
| Q4_K_M | 5.01 | Very good | 2.23 GB | 1.5 GB | 4.53 GB | 22.4 t/s | Offload |
| Q5_K_S | 5.53 | Very good | 2.46 GB | 1.5 GB | 4.76 GB | 20.3 t/s | Offload |
| Q5_K_M | 5.89 | Very good | 2.62 GB | 1.5 GB | 4.92 GB | 19.1 t/s | Offload |
| Q6_K | 6.57 | Excellent | 2.92 GB | 1.5 GB | 5.22 GB | 17.1 t/s | Offload |
| Q8_0 | 8.5 | Excellent | 3.78 GB | 1.5 GB | 6.08 GB | 13.2 t/s | Offload |
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 MaziyarPanahi/Phi-3.5-mini-instruct-GGUF?
You need about 5.22 GB of VRAM to run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF entirely on the GPU using the Q6_K quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF fully on the GPU using Q8_0 (about 6.08 GB).
Can I run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF fully on the GPU using Q8_0 (about 6.08 GB).
Can I run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run MaziyarPanahi/Phi-3.5-mini-instruct-GGUF fully on the GPU using Q8_0 (about 6.08 GB).
What is the best quantization for MaziyarPanahi/Phi-3.5-mini-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.