LLM EXPLORER 59,420 MODELS INDEXED

MaziyarPanahi Phi 3 Mini 4K Instruct V0.3 HQQ 1bit Smashed by PrunaAI

By PrunaAI · 5 downloads
HF Repo ↗ Copy URL

MaziyarPanahi Phi 3 Mini 4K Instruct V0.3 HQQ 1bit Smashed is an open-source language model by PrunaAI. Features: 4b LLM, VRAM: 0.9GB, Context: 4K, Quantized, Instruction-Based, LLM Explorer Score: 0.13.

  1bit Base model:finetune:maziyarpan... Base model:maziyarpanahi/calme...   Custom code   Instruct   Phi3   Pruna-ai   Quantized   Region:us
Model Card on HF 🤗: https://huggingface.co/PrunaAI/MaziyarPanahi-Phi-3-mini-4k-instruct-v0.3-HQQ-1bit-smashed 

MaziyarPanahi Phi 3 Mini 4K Instruct V0.3 HQQ 1bit Smashed Parameters and Internals

Model Type 
compressed
Use Cases 
Areas:
AI Model Compression
Applications:
Reducing resource usage in AI models
Primary Use Cases:
Improving AI model efficiency
Considerations:
Efficiency results may vary in other settings. It's recommended to test the model in specific use-case conditions.
Additional Notes 
This model uses safetensors format. First run metrics may vary due to CUDA overheads.
Training Details 
Methodology:
Compressed with hqq using WikiText as calibration data.
Input Output 
Performance Tips:
It's recommended to run efficiency tests directly in use-case scenarios to determine if the smashed model benefits your application.
LLM NameMaziyarPanahi Phi 3 Mini 4K Instruct V0.3 HQQ 1bit Smashed
Repository 🤗https://huggingface.co/PrunaAI/MaziyarPanahi-Phi-3-mini-4k-instruct-v0.3-HQQ-1bit-smashed 
Base Model(s)  MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3   MaziyarPanahi/Phi-3-mini-4k-instruct-v0.3
Model Size4b
Required VRAM0.9 GB
Updated2025-09-13
MaintainerPrunaAI
Model Typephi3
Instruction-BasedYes
Model Files  0.9 GB
Quantization Type1bit
Model ArchitecturePhi3ForCausalLM
Context Length4096
Model Max Length4096
Transformers Version4.40.0
Tokenizer ClassLlamaTokenizer
Padding Token<|endoftext|>
Vocabulary Size32064
Torch Data Typebfloat16

Best Alternatives to MaziyarPanahi Phi 3 Mini 4K Instruct V0.3 HQQ 1bit Smashed

Best Alternatives
Context / RAM
Downloads
Likes
Ruphi 4B128K / 7.6 GB60
Phi4 Mini Prose2Tags 4B128K / 7.7 GB200
Calme 2.1 Phi3.5 4B128K / 7.7 GB184
Vanta Research Apollo V1 4B128K / 7.7 GB70
AURORAV0.3 4B128K / 7.7 GB253
Lunar 4B128K / 7.7 GB51
Phi3.5 Gutenberg 4B128K / 7.7 GB134
Calme 2.1 Phi3 4B4K / 7.7 GB111
Calme 2.3 Phi3 4B4K / 7.7 GB669
Calme 2.2 Phi3 4B4K / 7.7 GB73
Note: green Score (e.g. "73.2") means that the model is better than PrunaAI/MaziyarPanahi-Phi-3-mini-4k-instruct-v0.3-HQQ-1bit-smashed.