LLM EXPLORER 60,645 MODELS INDEXED

Phi 3 Medium 4K Instruct 8bit by nold

By nold · 5 downloads
HF Repo ↗ Copy URL

Phi 3 Medium 4K Instruct 8bit is an open-source language model by nold. Features: 14b LLM, VRAM: 14.3GB, Context: 4K, License: mit, Quantized, Instruction-Based, LLM Explorer Score: 0.24, ELO: 1138, Arc: 67.3, HellaSwag: 85.8, MMLU: 77.8, GSM8K: 79.4.

  8-bit   8bit   Bitsandbytes   Code   Conversational   Custom code   Endpoints compatible   Instruct   Multilingual   Phi3   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/nold/phi-3-medium-4k-instruct-8bit 

Phi 3 Medium 4K Instruct 8bit Benchmarks

LMSys ELO 1138 vs 1272 (so35) -10.5%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

Phi 3 Medium 4K Instruct 8bit Parameters and Internals

Model Type 
text generation, instruction following
Use Cases 
Areas:
Commercial, Research
Applications:
General purpose AI systems, Memory/compute constrained environments, Latency bound scenarios, Strong reasoning (code, math, logic)
Primary Use Cases:
Intended for use in broad commercial and research fields
Limitations:
Not designed for all downstream purposes, Limited language support outside English
Considerations:
Evaluate and mitigate for accuracy, safety, and fairness
Supported Languages 
multilingual (English), others (10% multilingual)
Training Details 
Data Sources:
Publicly available documents, Newly created synthetic data, High quality chat format supervised data
Data Volume:
4.8T tokens
Methodology:
Supervised fine-tuning and Direct Preference Optimization (DPO)
Context Length:
4096
Training Time:
42 days
Hardware Used:
512 GPUs H100-80G
Model Architecture:
Dense decoder-only Transformer
Responsible Ai Considerations 
Fairness:
Awareness of language variety performance, representation of harms and stereotypes
Transparency:
Disclosures on potential behaviors
Accountability:
Developers are responsible
Mitigation Strategies:
Follow best practices and implement additional mitigations for sensitive deployment contexts
Input Output 
Input Format:
Chat-format
Accepted Modalities:
text
Output Format:
Text
Performance Tips:
Include specific tokens for improved reliability
LLM NamePhi 3 Medium 4K Instruct 8bit
Repository 🤗https://huggingface.co/nold/phi-3-medium-4k-instruct-8bit 
Base Model(s)  Phi 3 Medium 4K Instruct   microsoft/Phi-3-medium-4k-instruct
Model Size14b
Required VRAM14.3 GB
Updated2026-07-23
Maintainernold
Model Typephi3
Instruction-BasedYes
Model Files  4.8 GB: 1-of-3   5.0 GB: 2-of-3   4.5 GB: 3-of-3
Quantization Type8bit
Model ArchitecturePhi3ForCausalLM
Licensemit
Context Length4096
Model Max Length4096
Transformers Version4.41.1
Tokenizer ClassLlamaTokenizer
Padding Token<|endoftext|>
Vocabulary Size32064
Torch Data Typefloat16

Best Alternatives to Phi 3 Medium 4K Instruct 8bit

Best Alternatives
Context / RAM
Downloads
Likes
Phi 3 Medium 128K Instruct128K / 28 GB45634389
...ess V2.5 Phi 3 Medium 128K 14B128K / 28 GB164
...um 128K Instruct GPTQ 8 Bit C4128K / 14.6 GB60
...i 3 Medium 128K Instruct C4 32128K / 15.5 GB50
...edium 128K Instruct GPTQ 8 Bit128K / 14.6 GB50
Phi 3 Medium 128K Instruct AWQ128K / 7.8 GB141
Finetuned Phi3 Medium 128K128K / 28 GB80
Shisa V1 Phi3 14B128K / 28 GB52
Mahou 1.2 Phi 14B128K / 28 GB81
...colatine 14B Instruct DPO V1.316K / 29.4 GB160
Note: green Score (e.g. "73.2") means that the model is better than nold/phi-3-medium-4k-instruct-8bit.