Instructions to use prithivMLmods/biodecision-v2-4b-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use prithivMLmods/biodecision-v2-4b-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-classification", model="prithivMLmods/biodecision-v2-4b-GGUF")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("prithivMLmods/biodecision-v2-4b-GGUF", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use prithivMLmods/biodecision-v2-4b-GGUF 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 prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf prithivMLmods/biodecision-v2-4b-GGUF: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 prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf prithivMLmods/biodecision-v2-4b-GGUF: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 prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Use Docker

docker model run hf.co/prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use prithivMLmods/biodecision-v2-4b-GGUF with Ollama:
```
ollama run hf.co/prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
```
Unsloth Desktop

How to use prithivMLmods/biodecision-v2-4b-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use prithivMLmods/biodecision-v2-4b-GGUF with Docker Model Runner:
```
docker model run hf.co/prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M
```

Lemonade

How to use prithivMLmods/biodecision-v2-4b-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.biodecision-v2-4b-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use prithivMLmods/biodecision-v2-4b-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use prithivMLmods/biodecision-v2-4b-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "prithivMLmods/biodecision-v2-4b-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

biodecision-v2-4b-GGUF

BioDecision-4B v2 is a System-1 decision model for biomedicine, pharma and clinical trials, retrained from Qwen3.5-4B with a LoRA adapter (merged into the weights) using Together AI's Tev1 recipe. It reads a source, a question and a set of options (Choice with 2-24 options, Yes/No, or Score) and returns a calibrated probability for each answer in one forward pass, with no generated text. Training was one epoch on a 1,750,239-decision pool: all of v1's data, re-screened, plus 22 new sources covering drug and device safety, regulatory work, data standards, trial design, genomics, a biomedical knowledge graph and diagnosis. On the card's 63,432 held-out test decisions it scores 75.3%, up from 69.4% for v1 and ahead of TypeSafe's Jev 1.13 at 72.0%. On the 22 added sources it reaches 85.1% (v1: 68.4%). Against Jev 1.13 on 21 domain datasets it averages 73.9% versus 65.8%. Jev still leads on general medical knowledge exams (71.1% vs 61.2%, with MedQA at 74.6% for v2), on Chinese exams, and on the card's author-built 2025-26 cases (254 vs 212 of 284). It improves relation extraction (DrugProt F1 71.7, DDI-2013 68.6) and judging AI answers against a source (MedHallu 97.0%). It is weaker than v1 on three of the four CT Open trial-forecasting tasks and on TREC re-ranking, and the card tells users who need those to stay with v1. When it is at least 90% confident it was right 98.5% of the time on held-out dev decisions, covering 60% of answers, though the card notes these are observed rates and not guarantees. A bundled server offers a Jev-compatible /v1/systemone API on vLLM, with about 39 ms median latency on an A100. The weights are for research use only, since several training sources are non-commercial, and are not validated for patient care.

Model Files

File Name	Quant Type	File Size	File Link	Description
biodecision-v2-4b.BF16.gguf	BF16	8.42 GB	Link	Full BF16 weights. Highest quality, largest file size.
biodecision-v2-4b.Q3_K_L.gguf	Q3_K_L	2.42 GB	Link	Lower quality but usable, good for low RAM availability.
biodecision-v2-4b.Q3_K_M.gguf	Q3_K_M	2.26 GB	Link	Low quality.
biodecision-v2-4b.Q4_K_M.gguf	Q4_K_M	2.71 GB	Link	Good quality, default size for most use cases, recommended.
biodecision-v2-4b.Q4_K_S.gguf	Q4_K_S	2.56 GB	Link	Slightly lower quality with more space savings, recommended.
biodecision-v2-4b.Q5_K_M.gguf	Q5_K_M	3.07 GB	Link	High quality, recommended.
biodecision-v2-4b.Q5_K_S.gguf	Q5_K_S	2.99 GB	Link	High quality, recommended.
biodecision-v2-4b.Q6_K.gguf	Q6_K	3.46 GB	Link	Very high quality, near perfect, recommended.
biodecision-v2-4b.Q8_0.gguf	Q8_0	4.48 GB	Link	Extremely high quality, generally unneeded but max available quant.