LLM EXPLORER 59,516 MODELS INDEXED

Microsoft Phi 3 Mini 128K Instruct HQQ 2bit Smashed by PrunaAI

By PrunaAI · 8 downloads
HF Repo ↗ Copy URL

Microsoft Phi 3 Mini 128K Instruct HQQ 2bit Smashed is an open-source language model by PrunaAI. Features: LLM, VRAM: 1.4GB, Context: 128K, Quantized, Instruction-Based, LLM Explorer Score: 0.12.

  2bit Base model:finetune:microsoft/... Base model:microsoft/phi-3-min...   Custom code   Instruct   Phi3   Pruna-ai   Quantized   Region:us
Model Card on HF 🤗: https://huggingface.co/PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed 

Microsoft Phi 3 Mini 128K Instruct HQQ 2bit Smashed Parameters and Internals

Model Type 
compressed, optimized
Use Cases 
Limitations:
The quality of the model output might vary compared to the base model.
Considerations:
Efficiency results may vary in other settings. We recommend running them in the use-case conditions.
Additional Notes 
To compress your own models, contact PrunaAI for premium access and tech support for specific use-cases.
Training Details 
Methodology:
The model is compressed with hqq.
Input Output 
Input Format:
Tokens from text queries (e.g. 'What is the color of prunes?')
Accepted Modalities:
text
Output Format:
Text generation (decoded response)
Performance Tips:
Directly assess efficiency gains in your use-cases.
LLM NameMicrosoft Phi 3 Mini 128K Instruct HQQ 2bit Smashed
Repository 🤗https://huggingface.co/PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed 
Base Model(s)  Phi 3 Mini 128K Instruct   microsoft/Phi-3-mini-128k-instruct
Required VRAM1.4 GB
Updated2025-09-23
MaintainerPrunaAI
Model Typephi3
Instruction-BasedYes
Model Files  1.4 GB
Quantization Type2bit
Model ArchitecturePhi3ForCausalLM
Context Length131072
Model Max Length131072
Transformers Version4.42.4
Tokenizer ClassLlamaTokenizer
Padding Token<|endoftext|>
Vocabulary Size32064
Torch Data Typebfloat16

Best Alternatives to Microsoft Phi 3 Mini 128K Instruct HQQ 2bit Smashed

Best Alternatives
Context / RAM
Downloads
Likes
...m 128K Instruct 8.0bpw H8 EXL2128K / 13.4 GB74
...dium 128K Instruct 8 0bpw EXL2128K / 13.4 GB41
...m 128K Instruct 6.0bpw H6 EXL2128K / 10.7 GB73
...m 128K Instruct 3.0bpw H6 EXL2128K / 5.6 GB50
...m 128K Instruct 5.0bpw H6 EXL2128K / 8.9 GB50
...28K Instruct Ov Fp16 Int4 Asym128K / 2.5 GB90
...128K Instruct HQQ 4bit Smashed128K / 2.9 GB70
NuExtract Bpw6 EXL24K / 3 GB41
...Mini 4K Geminified 3 0bpw EXL24K / 1.6 GB60
...uct Abliterated V3 2 2bpw EXL24K / 4.2 GB90
Note: green Score (e.g. "73.2") means that the model is better than PrunaAI/microsoft-Phi-3-mini-128k-instruct-HQQ-2bit-smashed.