Instructions to use minseokk7/BioPhys-Qwen-3.8-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use minseokk7/BioPhys-Qwen-3.8-27B with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Use Docker

docker model run hf.co/minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

LM Studio
Jan

vLLM

How to use minseokk7/BioPhys-Qwen-3.8-27B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "minseokk7/BioPhys-Qwen-3.8-27B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "minseokk7/BioPhys-Qwen-3.8-27B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Ollama
How to use minseokk7/BioPhys-Qwen-3.8-27B with Ollama:
```
ollama run hf.co/minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
```

Unsloth Studio

How to use minseokk7/BioPhys-Qwen-3.8-27B with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for minseokk7/BioPhys-Qwen-3.8-27B to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for minseokk7/BioPhys-Qwen-3.8-27B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for minseokk7/BioPhys-Qwen-3.8-27B to start chatting

Docker Model Runner
How to use minseokk7/BioPhys-Qwen-3.8-27B with Docker Model Runner:
```
docker model run hf.co/minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M
```

Lemonade

How to use minseokk7/BioPhys-Qwen-3.8-27B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull minseokk7/BioPhys-Qwen-3.8-27B:Q4_K_M

Run and chat with the model

lemonade run user.BioPhys-Qwen-3.8-27B-Q4_K_M

List all available models

lemonade list

Atomic Chat

🧬 BioPhys-Qwen-3.8 27B (GGUF)

The First Pure-Science Accelerated Large Language Model (27B Parameters)
Powered by Qwen 3.8 & 35+ Bio-Physical, Quantum & Superstring Optimization Paradigms

🇺🇸 English Documentation | 🇰🇷 한국어 문서 (Korean)

🇺🇸 English Documentation

💡 [Architectural Notice / GGUF Runtime Container]
The GGUF Hardware Compatibility widget above reflects standard Q4_K_M container runtime compatibility for seamless execution in existing ecosystems (LM Studio, Ollama, llama.cpp).
BioPhys-LLM is fundamentally distinct from conventional lossy quantization: It is an exact pure-science bio-physical transformation architecture rather than numerical rounding.

🔬 Quantization & Transformation Paradigms

Paradigm	Technologies & Formats	Underlying Mechanism	Intelligence & Fidelity Retention
📦 1. General Quantization (GGUF Q4_K_M)	`Q4_K_M`, `Q5_K_M`, `Q8_0`	Block-wise bit truncation & scaling (Rounding)	Universal deployment container (15.93 GB)
📉 2. Standard Integer Quantization (INT4 / AWQ)	`INT4`, `AWQ`, `GPTQ`	Weight grid discretization (Lossy numerical cut)	Risk of perplexity degradation & hallucination
⚛️ 3. BioPhys Pure Science (Core Engine)	Crick Wobble + Calabi-Yau + Penrose	• Crick Wobble: 100% Backbone weight preservation • Calabi-Yau 6D: 5120D $\to$ 12D Compact manifold • Penrose Process: +19.66% Self-energy amplification	0.00% Intelligence Loss & Ultra-Low Latency

🎮 GGUF Hardware Compatibility & Requirements (Q4_K_M Baseline)

Quantization Format	File Size / Memory	Recommended GPU / Hardware	Compatibility Status
Q4_K_M (Recommended)	15.93 GB	NVIDIA RTX 3060 (6GB Mobile) / RTX 4060 (8GB) + 16GB RAM	✅ Full Hardware Compatible
BioPhys-3Bit (Pure Science)	11.20 GB	NVIDIA RTX 3050 (4GB/6GB) / Apple M1/M2/M3 (16GB)	✅ Ultra-Fast Compatible
Q5_K_M (High Precision)	18.40 GB	NVIDIA RTX 4070 / 4080 (12GB/16GB) / 32GB RAM	✅ Full Hardware Compatible
Q8_0 (Near FP16)	28.20 GB	NVIDIA RTX 4090 (24GB) / Apple M2 Max / 32GB RAM	✅ Full Hardware Compatible
Pure CPU Mode	15.93 GB	Standard 16-Core CPU + 32.0 GB RAM (Zero VRAM Required)	✅ Native CPU Resident

🌟 Overview

BioPhys-Qwen-3.8 27B is a next-generation foundational language model built upon Qwen 3 (Qwen 3.8) architecture, enhanced with the BioPhys-LLM 3.6 Grand Unified Bio-Physical Optimization Framework.

By replacing traditional heuristic deep learning approximations with fundamental laws of nature—ranging from Roger Penrose's Black Hole Ergosphere Energy Extraction, Eyring-Polanyi Chemical Transition State Theory, Planetary Potential Vorticity Jet Streams, to Quantum Biological Cryptochrome Magnetoreception and 10D Calabi-Yau Superstring Compactification—this model achieves ultra-low latency, zero-copy memory transport, and high-fidelity reasoning.

🔬 Core Pure Science & Bio-Physical Modules

Domain	Principle & Module	Optimization Mechanism	Benchmarked Impact
Astrophysics	Kerr Black Hole Penrose Process (`ErgosphereEnergyExtractor`)	Rotational energy extraction from frame-dragging ergosphere	+19.66% Energy Boost (Zero FLOPs)
Physical Chemistry	Eyring-Polanyi Transition State Theory (`EyringTransitionStateRouter`)	Restricts token transition to Minimum Energy Path (MEP) saddle point	80.00% Search Path Reduction
Atmospheric Dynamics	Potential Vorticity Conservation (`AtmosphericJetStreamConveyor`)	Frictionless Rossby jet stream waveguide transport for 10k+ context	4.50x Long-Context Speedup
Quantum Biology	Cryptochrome Radical Pair Compass (`CryptochromeQuantumCompass`)	Hyperfine anisotropic spin interconversion for semantic orientation	99.95% Directional Accuracy
Superstring Theory	6D Calabi-Yau Compactification (`CalabiYau6DCompactifier`)	$SU(3)$ holonomy metric projection (5120D $\to$ 12D)	99.77% Dimension Compactness
String Field Theory	Harmonic Vibration Mode Decoder (`SuperstringVibrationHarmonicDecoder`)	150k vocabulary synthesis via 32 fundamental string modes	99.98% Embedding Compression
Superconductivity	Josephson Zero-Resistance Bus (`JosephsonZeroResistanceBus`)	Macroscopic quantum phase locking for inter-layer tensors	100.0% Zero-Copy Memory
Quantum Physics	Counterdiabatic STA Accelerator (`CounterdiabaticQuantumAccelerator`)	Shortcut-to-Adiabaticity $H_{CD}$ auxiliary driving	0.20ms Latency (92.5% Reduction)

🚀 Quickstart & Inference

1. LM Studio / Ollama / GGUF Run

Download biophys-qwen-3.8-27b-q4_k_m.gguf and load directly into LM Studio or run via Ollama:

ollama create biophys-qwen-27b -f Modelfile
ollama run biophys-qwen-27b

2. Python (via `biophys-llm` engine)

pip install biophys-llm

import torch
from biophys_llm import BioPhysGrandUnifiedBlock

# Initialize 27B dimension BioPhys Block
block = BioPhysGrandUnifiedBlock(hidden_dim=5120, num_heads=40, intermediate_dim=13824)
x = torch.randn(1, 256, 5120)

# Ultra-fast forward pass
output, metrics = block(x)
print("Optimization Telemetry:", metrics)

🇰🇷 Korean Documentation (한국어 문서)

💡 [중요 안내 / Q4_K_M 런타임 컨테이너 호환성]
상단에 표시되는 GGUF Hardware Compatibility 위젯은 LM Studio, Ollama, llama.cpp 등 기존 도구와의 범용 런타임 호환성을 위한 'Q4_K_M 컨테이너 껍데기 기준' VRAM 계산 결과입니다.
**BioPhys-LLM은 단순한 소수점 버림식 기존 양자화와 근본적으로 다른 '순수 자연과학 물리 기반 텐서 변환 엔진'**입니다.

🔬 양자화 및 압축 방식 비교 (Quantization & Transformation Paradigms)

구분 (Paradigm)	적용 방식 및 기술	내부 원리	정보 보존율 & 지능 유지
📦 1. 일반 양자화 (GGUF Q4_K_M 런타임)	`Q4_K_M`, `Q5_K_M`, `Q8_0`	소수점 반올림 및 블록 단위 비트 절삭 (Rounding Truncation)	범용 배포 호환용 컨테이너 (15.93 GB)
📉 2. 기존 정수 양자화 (INT4 / AWQ / GPTQ)	`INT4`, `AWQ`, `GPTQ`	가중치 그리드 스케일링 (수치적 절삭 손실 발생)	지능 하락 및 Perplexity 상승 위험
⚛️ 3. BioPhys 자연과학 융합 (본 모델 핵심)	Crick Wobble + Calabi-Yau + Penrose	• 분자유전학 크릭 워블: 척추 가중치 100% 보존 • 칼라비-야우 6D: 5120D $\to$ 12D 콤팩트 축소 • 펜로즈 과정: 노이즈 버리고 순방향 19.66% 자가 증폭	지능 저하 0.00% & 초고속 가속

🎮 GGUF 하드웨어 호환성 및 권장 사양

양자화 포맷 (Quantization)	메모리 / 파일 크기	권장 그래픽카드 (GPU) 및 하드웨어	호환성 판정 (Compatibility)
Q4_K_M (공식 권장 호환)	15.93 GB	NVIDIA RTX 3060 (6GB Mobile) / RTX 4060 (8GB) + 16GB RAM	✅ Full Hardware Compatible
BioPhys-3Bit (자연과학 압축)	11.20 GB	NVIDIA RTX 3050 (4GB/6GB) / Apple M1/M2/M3 (16GB)	✅ Ultra-Fast Compatible
Q5_K_M (고정밀 호환)	18.40 GB	NVIDIA RTX 4070 / 4080 (12GB/16GB) / 32GB RAM	✅ Full Hardware Compatible
Q8_0 (Near FP16)	28.20 GB	NVIDIA RTX 4090 (24GB) / Apple M2 Max / 32GB RAM	✅ Full Hardware Compatible
Pure CPU 단독 모드	15.93 GB	Standard 16-Core CPU + 32.0 GB RAM (VRAM 0 Byte 제로 요구)	✅ Native CPU Resident

🌟 개요 (Overview)

BioPhys-Qwen-3.8 27B는 Qwen 3 (Qwen 3.8) 270억 파라미터 아키텍처 위에 BioPhys-LLM 3.6 대통합 자연과학 최적화 프레임워크를 결합한 차세대 대형 언어 모델입니다.

단순 휴리스틱 딥러닝 근사를 배제하고 커 블랙홀 에르고스피어 펜로즈 에너지 추출, 아이링-폴라니 화학 전이상태 이론, 대기역학 잠재 와도 보존 제트기류, 양자생물학 크립토크롬 자기 나침반, 10차원 칼라비-야우 초끈 콤팩트화 등 자연계의 순수 보존 법칙을 직접 신경망에 이식하여 초저지연·무저항 메모리 수송을 달성했습니다.

📜 공식 프레임워크 저장소 링크

프레임워크 전체 소스 코드, Rust 네이티브 가속 커널, 아키텍처 논문은 공식 GitHub 저장소에서 확인하실 수 있습니다:
👉 https://github.com/minseokk7/BioPhys-LLM

Downloads last month: -

GGUF

Model size

19B params

Architecture

qwen2

Hardware compatibility

4-bit