---

base_model: Qwen/Qwen3-0.6B

library_name: gguf

pipeline_tag: text-generation

tags:

- gguf

- llama.cpp

- quantized

- imatrix

---

Qwen3-0.6B GGUF

GGUF quantizations of Qwen/Qwen3-0.6B, covering 30 files (12.8 GB total).

Files

| File | Quant | Size | Notes |

|---|---|---:|---|

| Qwen3-0.6B-BF16.gguf | BF16 | 1.41 GB | Full precision source. Every quant below is cut from this file. |

| Qwen3-0.6B-Q8_0.gguf | Q8_0 | 0.75 GB | Effectively lossless. Use when disk and RAM are not the constraint. |

| Qwen3-0.6B-Q6_K.gguf | Q6_K | 0.58 GB | Near-lossless; the last stop before quality becomes measurable. |

| Qwen3-0.6B-Q5_K_M.gguf | Q5_K_M | 0.51 GB | Very good quality, noticeably smaller than Q6_K. |

| Qwen3-0.6B-Q5_K_S.gguf | Q5_K_S | 0.51 GB | Slightly smaller than Q5_K_M for a slight quality cost. |

| Qwen3-0.6B-Q5_1.gguf | Q5_1 | 0.54 GB | Legacy. Prefer Q5_K_M. |

| Qwen3-0.6B-Q5_0.gguf | Q5_0 | 0.51 GB | Legacy. Prefer Q5_K_M. |

| Qwen3-0.6B-Q4_K_M.gguf | Q4_K_M | 0.45 GB | The usual default. Best quality-per-byte for most people. |

| Qwen3-0.6B-Q4_K_S.gguf | Q4_K_S | 0.44 GB | A little smaller than Q4_K_M, a little worse. |

| Qwen3-0.6B-IQ4_NL.gguf | IQ4_NL | 0.44 GB | Non-linear 4-bit; good on hardware without fast K-quant kernels. |

| Qwen3-0.6B-IQ4_XS.gguf | IQ4_XS | 0.42 GB | Best sub-4.5bpw option; usually beats Q4_K_S at a smaller size. |

| Qwen3-0.6B-Q4_1.gguf | Q4_1 | 0.47 GB | Legacy. Prefer Q4_K_M. |

| Qwen3-0.6B-Q4_0.gguf | Q4_0 | 0.44 GB | Legacy round-to-nearest. Prefer Q4_K_M unless a runtime needs this. |

| Qwen3-0.6B-Q3_K_L.gguf | Q3_K_L | 0.41 GB | Small, with real quality loss. Usable when RAM is tight. |

| Qwen3-0.6B-Q3_K_M.gguf | Q3_K_M | 0.39 GB | Smaller again; noticeable degradation. |

| Qwen3-0.6B-IQ3_M.gguf | IQ3_M | 0.38 GB | Strong at ~3.7bpw, clearly better than Q3_K_M. |

| Qwen3-0.6B-IQ3_S.gguf | IQ3_S | 0.36 GB | Slightly smaller than IQ3_M. |

| Qwen3-0.6B-Q3_K_S.gguf | Q3_K_S | 0.36 GB | Aggressive. Prefer IQ3_M at a similar size. |

| Qwen3-0.6B-IQ3_XS.gguf | IQ3_XS | 0.35 GB | Aggressive but coherent. |

| Qwen3-0.6B-IQ3_XXS.gguf | IQ3_XXS | 0.32 GB | Very aggressive; imatrix carries it. |

| Qwen3-0.6B-Q2_K.gguf | Q2_K | 0.32 GB | Very small, heavily degraded. For experimentation. |

| Qwen3-0.6B-IQ2_M.gguf | IQ2_M | 0.31 GB | The smallest size most people find usable. |

| Qwen3-0.6B-Q2_K_S.gguf | Q2_K_S | 0.31 GB | Smaller than Q2_K, requires the imatrix. |

| Qwen3-0.6B-IQ2_S.gguf | IQ2_S | 0.30 GB | Below the usual usability line. |

| Qwen3-0.6B-IQ2_XS.gguf | IQ2_XS | 0.27 GB | Experimental. |

| Qwen3-0.6B-IQ2_XXS.gguf | IQ2_XXS | 0.26 GB | Experimental. |

| Qwen3-0.6B-Q2_0.gguf | Q2_0 | 0.32 GB | Extreme, group-64. Included for completeness. |

| Qwen3-0.6B-IQ1_M.gguf | IQ1_M | 0.25 GB | Extreme. Expect substantial degradation. |

| Qwen3-0.6B-IQ1_S.gguf | IQ1_S | 0.24 GB | Extreme. Expect substantial degradation. |

| Qwen3-0.6B-Q1_0.gguf | Q1_0 | 0.20 GB | Extreme. Included for completeness. |

Which one should I download?

Pick the largest file that leaves a couple of gigabytes of headroom on the device you will run it on — the model has to fit in RAM (or VRAM, if you are offloading) alongside the KV cache and the OS.

• Plenty of memory: Q6_K or Q8_0.
• The usual choice: Q4_K_M.
• Tight on memory: IQ4_XS, then IQ3_M, then IQ2_M.
• The IQ files are imatrix-guided and generally beat a Q file of similar size, at the cost of slightly slower inference on some hardware.

Quantization details

• Importance matrix computed with llama-imatrix over 500 rows of Salesforce/wikitext (wikitext-2-raw-v1).
• The matrix was computed on the BF16 weights.
• K-quants below 6 bit and the whole IQ set are imatrix-guided. Q4_0/Q4_1/Q5_0/Q5_1 are legacy round-to-nearest and ignore it; Q6_K/Q8_0 are near-lossless and do not need it.
• All files are cut from the same BF16 GGUF, so differences between them are quantization only.

Usage

llama-cli -hf NANI-Nithin/Qwen3-0.6B-GGUF:Q4_K_M -p "Hello"

Or download one file and point at it directly:

huggingface-cli download NANI-Nithin/Qwen3-0.6B-GGUF Qwen3-0.6B-Q4_K_M.gguf --local-dir .
llama-cli -m Qwen3-0.6B-Q4_K_M.gguf -p "Hello"

---

Quantized with llama.cpp by AgentQuantix on 2026-09-04. Licensing follows the base model.