Instructions to use frontterrain/HARP-1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use frontterrain/HARP-1B 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 frontterrain/HARP-1B:F16
# Run inference directly in the terminal:
llama cli -hf frontterrain/HARP-1B:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf frontterrain/HARP-1B:F16
# Run inference directly in the terminal:
llama cli -hf frontterrain/HARP-1B:F16

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 frontterrain/HARP-1B:F16
# Run inference directly in the terminal:
./llama-cli -hf frontterrain/HARP-1B:F16

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 frontterrain/HARP-1B:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf frontterrain/HARP-1B:F16

Use Docker

docker model run hf.co/frontterrain/HARP-1B:F16

Llama new
LM Studio
Jan

vLLM

How to use frontterrain/HARP-1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "frontterrain/HARP-1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "frontterrain/HARP-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/frontterrain/HARP-1B:F16

Ollama
How to use frontterrain/HARP-1B with Ollama:
```
ollama run hf.co/frontterrain/HARP-1B:F16
```
Unsloth Desktop
Docker Model Runner
How to use frontterrain/HARP-1B with Docker Model Runner:
```
docker model run hf.co/frontterrain/HARP-1B:F16
```

Lemonade

How to use frontterrain/HARP-1B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull frontterrain/HARP-1B:F16

Run and chat with the model

lemonade run user.HARP-1B-F16

List all available models

lemonade list

Atomic Chat

HARP-1B (FrontTerrain Foundation Model)

HARP-1B is a high-performance, compact autoregressive language model designed and trained from scratch by FrontTerrain.

It is specifically optimized for systems programming, mathematics, Python, and native FrontLang (.ft) source code generation.

🏛️ FrontTerrain Leadership & Core Team

Founder & CEO: Rishi Shandilya (@dark2quasar)
AI Lead: Arka Brahma
CTO: Tanishk Tripathi
COO: Abhinay Singh
CMO: Atharva Srivastava

📊 Model Architecture & Training Specs

Specification	Value
Parameters	98,913,024 (~99M Params)
Architecture	Rotary Position Embeddings (RoPE), SwiGLU, RMSNorm, Grouped-Query Attention (GQA)
Layers	12
Hidden Dimension	768
Attention Heads	12 Query Heads, 3 Key-Value Heads (4:1 GQA)
Head Dimension	64
Vocabulary Size	32,000
Context Length	512 tokens
Precision	`bfloat16` Native Mixed-Precision
Acceleration	NVIDIA L4 Tensor Cores + FlashAttention + PyTorch Triton Inductor JIT
Current Training Step	52,000
Trained Tokens	1,400,832,000 Tokens (~1.40 Billion)
Target Tokens	50,000,000,000 (50 Billion Tokens)
Training Speed	~40,800 tokens/sec
Current Training Loss	0.2773

🌟 Native FrontLang (`.ft`) & Ecosystem Support

HARP-1B is pre-trained with comprehensive documentation and code for the FrontTerrain developer ecosystem:

FrontLang (.ft): Native compiler syntax, memory safety rules, structs, traits, pattern matching, bytecode specifications.
TensorLib: High-performance tensor computation and deep learning framework.
Codrop: Collaborative modern IDE ecosystem.

Example FrontLang Code Output:

// FrontLang Matrix Multiplication Example
struct Matrix {
    rows: usize,
    cols: usize,
    data: [f32]
}

fn main() {
    println("Hello from FrontLang running on HARP-1B!");
}

🚀 Quickstart Inference (Python)

import torch
from safetensors.torch import load_file

# 1. Load weights
weights = load_file("model.safetensors")
print(f"Loaded HARP-1B with {len(weights)} weight tensors!")

📜 License

Apache License 2.0. Developed with ❤️ by FrontTerrain.

Downloads last month: -

Safetensors

Model size

0.1B params

Tensor type

BF16