How to use from
llama.cpp
Install from brew
brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf large-traversaal/Alif-1.0-3B-Instruct: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 large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf large-traversaal/Alif-1.0-3B-Instruct: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 large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
Use Docker
docker model run hf.co/large-traversaal/Alif-1.0-3B-Instruct:Q4_K_M
Quick Links

Alif 1.0 3B Instruct

Alif 1.0 3B Instruct is an open-source instruction-tuned language model developed by traversaal.ai, focused on Urdu and English understanding and reasoning. Fine-tuned with high-quality, culturally aware synthetic data, Alif delivers strong multilingual performance with a particular emphasis on low-resource language alignment and nuanced instruction following.


Key Highlights

  • Multilingual: Optimized for Urdu (primary) and English (secondary)
  • Instruction-tuned: Trained on synthetic QA, chat, and reasoning tasks
  • Lightweight: Only 3 billion parameters — fast, memory-efficient
  • 4-bit Quantized Version: Available for on-device inference
  • Training Framework: Fine-tuned 2× faster using Unsloth + TRL

Model Metadata


Use Cases

Alif is ideal for:

  • Urdu and English chatbots
  • Question answering in low-resource languages
  • Translation, summarization, and creative writing
  • Running on edge devices with the 4-bit version

Citation

@article{ShafiqueAlif2025,
  title        = {Alif: Advancing Urdu Large Language Models via Multilingual Synthetic Data Distillation},
  author       = {Muhammad Ali Shafique and Kanwal Mehreen and Muhammad Arham and Maaz Amjad and Sabur Butt and Hamza Farooq},
  journal      = {arXiv preprint arXiv:2510.09051},
  year         = {2025},
  url          = {https://arxiv.org/abs/2510.09051}
}
Downloads last month
50
GGUF
Model size
4B params
Architecture
llama
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for large-traversaal/Alif-1.0-3B-Instruct

Quantized
(54)
this model

Paper for large-traversaal/Alif-1.0-3B-Instruct