Skip to content
localllm
日本語

huihui-ai

Qwen2.5-3B-Instruct-abliterated

Parameters
3.4B
Max context length
32,768
Architecture
qwen2
License
Commercial OK
Base model
huihui-ai/Qwen2.5-3B-Instruct-abliterated
License
apache-2.0

About Qwen2.5-3B-Instruct-abliterated

Qwen2.5-3B-Instruct-abliterated is a GGUF model based on huihui-ai/Qwen2.5-3B-Instruct-abliterated on Hugging Face. GGUF files are published by mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF. It has about 3.4B parameters, the qwen2 architecture, a maximum context length of 32,768 tokens (values written in the GGUF).

There are 12 quantizations, with file sizes from 1.27 GB (Q2_K) to 6.80 GB (F16) (actual sizes on Hugging Face).

Running Q4_K_M (1.93 GB) with an 8,192-token context needs an estimated 3.31 GB of memory, and an estimated 4.21 GB at 32,768 tokens. For 8,192 tokens, a GPU with 4 GB or more of VRAM (such as Arc A310, Arc A370M), or a Mac with 4 GB or more of unified memory, is a rough guide (macOS caps how much memory the GPU can use, so leave some headroom). The smallest, Q2_K (1.27 GB), needs an estimated 2.65 GB at 8,192 tokens.

According to mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF, the license is Apache License 2.0, which allows commercial use (keep the license text and copyright notice when redistributing). Check the original license for the terms of use.

File size and estimated memory per quantization

  • Up to 8 GB
  • Up to 16 GB
  • Up to 24 GB
  • Up to 48 GB
  • Up to 128 GB
  • More

File sizes are the actual values on Hugging Face. Memory is estimated for each context length (KV cache f16). The formula is in How memory is calculated.

All memory figures are Estimated (KV cache f16). It should run on a setup with at least this much VRAM (or Mac unified memory). Click a value to see the devices and models that fit that amount (Find by GPU / Mac).

About the "quality guide" (how far a quantization drifts from the original)

For each quantization type, how far the output drifts from the original model, grouped into six levels from values llama.cpp measured on Llama-3-8B.

The value is mean KLD (the difference between the original and quantized model in next-token probabilities); smaller means closer to the original. Quantizations made with an imatrix (importance data) come out slightly smaller. Levels use the value without an imatrix (or with one, for types that only have that).

These values were not measured on this model. Larger models and MoE models tend to drift less with the same type. Distributor-specific mixed types (such as UD-Q4_K_XL) cannot be matched by type name alone, so they show "-".

Quantization Quality guide KLD (no imatrix) KLD (imatrix) Size (Llama-3-8B)
Q8_0 Nearly identical 0.0014 - 7.96 GiB
Q6_K Very small 0.0055 - 6.14 GiB
Q5_K_M Very small 0.0108 - 5.33 GiB
Q5_K_S Small 0.0166 - 5.21 GiB
Q5_1 Small 0.0180 - 5.65 GiB
Q5_0 Small 0.0222 - 5.21 GiB
Q4_K_M Small 0.0313 0.0282 4.58 GiB
Q4_K_S Medium 0.0431 0.0320 4.37 GiB
IQ4_NL Small - 0.0357 4.35 GiB
IQ4_XS Small - 0.0363 4.14 GiB
Q4_1 Medium 0.0717 - 4.78 GiB
Q4_0 Medium 0.0719 - 4.34 GiB
Q3_K_L Medium 0.0872 0.0731 4.03 GiB
Q3_K_M Medium 0.1019 0.0844 3.74 GiB
IQ3_M Medium - 0.1025 3.53 GiB
IQ3_S Medium - 0.1113 3.42 GiB
IQ3_XS Large - 0.1387 3.28 GiB
IQ3_XXS Large - 0.1836 3.05 GiB
Q3_K_S Large 0.2282 0.1998 3.41 GiB
IQ2_M Large - 0.3260 2.74 GiB
Q2_K Very large 0.4451 0.3322 2.96 GiB
Q2_K_S Very large - 0.4034 2.96 GiB
IQ2_S Very large - 0.4392 2.56 GiB
IQ2_XS Very large - 0.5463 2.43 GiB
IQ2_XXS Very large - 0.8120 2.24 GiB
IQ1_M Very large - 1.3931 2.01 GiB
IQ1_S Very large - 2.2546 1.88 GiB

Source: llama.cpp (MIT), measured on Llama-3-8B against FP16, Wikitext-2, revision f364eb6f. Original table · Reading and choosing quantizations

GGUF source

mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF

Converted and published by a third party Modified version (name includes abliterated etc.) About this label

Using the original model requires a request or agreement on Hugging Face (this does not affect downloading the GGUF files).

Last checked: Nov 11, 2026

Architecture
qwen2
Supported before llama.cpp b4409
Parameters
3.4B
Max context length
32,768
License
apache-2.0 Commercial OK
File size and memory per quantization for Qwen2.5-3B-Instruct-abliterated (mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 max
Q2_K Qwen2.5-3B-Instruct-abliterated.Q2_K~75a3ed91 1.27 GB Very large 2.50 GB 2.65 GB 2.95 GB 3.56 GB
Q2_K Qwen2.5-3B-Instruct-Abliterated.Q2_K 1.38 GB Very large 2.60 GB 2.75 GB 3.05 GB 3.66 GB
Q3_K_S Qwen2.5-3B-Instruct-abliterated.Q3_K_S~bfab29e2 1.45 GB Large 2.68 GB 2.83 GB 3.13 GB 3.74 GB
Q3_K_S Qwen2.5-3B-Instruct-Abliterated.Q3_K_S 1.59 GB Large 2.81 GB 2.96 GB 3.27 GB 3.87 GB
Q3_K_M Qwen2.5-3B-Instruct-abliterated.Q3_K_M~2891a8e3 1.59 GB Medium 2.82 GB 2.97 GB 3.27 GB 3.87 GB
Q3_K_L Qwen2.5-3B-Instruct-abliterated.Q3_K_L~0807b1af 1.71 GB Medium 2.93 GB 3.08 GB 3.39 GB 3.99 GB
Q3_K_M Qwen2.5-3B-Instruct-Abliterated.Q3_K_M 1.72 GB Medium 2.95 GB 3.10 GB 3.40 GB 4.01 GB
IQ4_XS Qwen2.5-3B-Instruct-abliterated.IQ4_XS~0bfd719c 1.75 GB Small 2.98 GB 3.13 GB 3.43 GB 4.03 GB
Q4_K_S Qwen2.5-3B-Instruct-abliterated.Q4_K_S~ff9a9ee5 1.83 GB Medium 3.06 GB 3.21 GB 3.51 GB 4.12 GB
Q3_K_L Qwen2.5-3B-Instruct-Abliterated.Q3_K_L 1.84 GB Medium 3.07 GB 3.22 GB 3.52 GB 4.12 GB
IQ4_XS Qwen2.5-3B-Instruct-Abliterated.IQ4_XS 1.92 GB Small 3.14 GB 3.29 GB 3.60 GB 4.20 GB
Q4_K_M Qwen2.5-3B-Instruct-abliterated.Q4_K_M~fa3c818a 1.93 GB Small 3.15 GB 3.31 GB 3.61 GB 4.21 GB
Q4_K_S Qwen2.5-3B-Instruct-Abliterated.Q4_K_S 2.01 GB Medium 3.23 GB 3.39 GB 3.69 GB 4.29 GB
Q4_K_M Qwen2.5-3B-Instruct-Abliterated.Q4_K_M 2.10 GB Small 3.33 GB 3.48 GB 3.78 GB 4.39 GB
Q5_K_S Qwen2.5-3B-Instruct-abliterated.Q5_K_S~5448c43d 2.17 GB Small 3.39 GB 3.55 GB 3.85 GB 4.45 GB
Q5_K_M Qwen2.5-3B-Instruct-abliterated.Q5_K_M~66486a94 2.22 GB Very small 3.45 GB 3.60 GB 3.90 GB 4.51 GB
Q5_K_S Qwen2.5-3B-Instruct-Abliterated.Q5_K_S 2.38 GB Small 3.61 GB 3.76 GB 4.06 GB 4.67 GB
Q5_K_M Qwen2.5-3B-Instruct-Abliterated.Q5_K_M 2.44 GB Very small 3.66 GB 3.81 GB 4.12 GB 4.72 GB
Q6_K Qwen2.5-3B-Instruct-abliterated.Q6_K~dab63ca0 2.54 GB Very small 3.76 GB 3.91 GB 4.22 GB 4.82 GB
Q6_K Qwen2.5-3B-Instruct-Abliterated.Q6_K 2.79 GB Very small 4.02 GB 4.17 GB 4.47 GB 5.08 GB
Q8_0 Qwen2.5-3B-Instruct-abliterated.Q8_0~3affec4d 3.29 GB Nearly identical 4.51 GB 4.66 GB 4.96 GB 5.57 GB
Q8_0 Qwen2.5-3B-Instruct-Abliterated.Q8_0 3.62 GB Nearly identical 4.84 GB 4.99 GB 5.29 GB 5.90 GB
F16 6.80 GB Original precision 8.03 GB 8.18 GB 8.48 GB 9.08 GB 
Ollama  ollama run hf.co/mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF:Q4_K_M
llama.cpp  llama-server -hf mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF:Q4_K_M

Generation speed on common GPUs and Macs (estimated)

A rough guide to how fast Q4_K_M (1.93 GB) generates text. Each token reads the model weights once, so the upper limit is memory bandwidth ÷ file size. The guide is 60–80% of that, where most published measurements fall.

Device Memory bandwidth Guide (tokens/s) Estimated Upper limit
Arc A370M 4 GB 112 GB/s 35–46 58
Arc A770 (8GB) 8 GB 512 GB/s 159–212 265
GeForce RTX 5090 32 GB 1,792 GB/s 557–743 929
Apple M2 8 GB 100 GB/s 31–41 52
Apple M2 Pro 16 GB 200 GB/s 62–83 104
Apple M5 Ultra 96 GB 1,200 GB/s 373–497 622

Estimates only. Prompt processing speed (limited by compute) is not included. Longer contexts and offloading part of the model to the CPU make it slower. Macs are shown only when installed memory is at least 4/3 of what is needed. Devices without published bandwidth (some GeForce cards, etc.) are not shown.

License and commercial use

Apache License 2.0 Commercial OK

Allows commercial use, modification and redistribution. When redistributing, keep the license text and copyright notice, and note changes in modified files.

The original model's license has not been fetched yet, so this is based on the GGUF repository's label. Original license text (checked Oct 18, 2026)

The summary states facts this site read from the original text and is not legal advice. Always check the original license and its latest terms before use. All licenses

How to run

Example commands using Q4_K_M. To use another quantization, replace the name with one from the table above. How to run local LLMs

Ollama

Run it by pointing directly at the GGUF on Hugging Face.

ollama run hf.co/mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF:Q4_K_M

llama.cpp

Pass the repository and quantization name to llama-server (or llama-cli) with -hf.

llama-server -hf mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF:Q4_K_M

LM Studio

Paste mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF into the search box, download it, and choose the quantization from the file list.

Frequently asked questions

How much memory do I need to run Qwen2.5-3B-Instruct-abliterated (Q4_K_M)?
The file size is 1.93 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 3.31 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Can Qwen2.5-3B-Instruct-abliterated be used commercially?
The license is Apache License 2.0 (Commercial use allowed).Allows commercial use, modification and redistribution. When redistributing, keep the license text and copyright notice, and note changes in modified files. Check the original license text before deciding.
Does Qwen2.5-3B-Instruct-abliterated run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?
8 GB: up to F16 (est. 8.18 GB); 12 GB: up to F16 (est. 8.18 GB); 16 GB: up to F16 (est. 8.18 GB); 24 GB: up to F16 (est. 8.18 GB) (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run Qwen2.5-3B-Instruct-abliterated with Ollama?
Run "ollama run hf.co/mradermacher/Qwen2.5-3B-Instruct-abliterated-GGUF:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.