nioncorp/nil-3b-gguf

nil-3b (1.8GB GGUF Soveriegn Model)

nil-3b is a sovereign, offline-first 3-billion parameter language model quantized down to a 1.8GB GGUF footprint (q4_k_m). Developed by Nion Technologies, Inc., nil-3b is engineered specifically for low-resource environments, zero-bandwidth deployments, and edge computing hardware (ARM Cortex-A72, wearables, and low-power devices) across Africa.

Model Highlights

• Footprint: 1.8GB (q4_k_m GGUF quantization)
• Memory Overhead: <120MB active operational RAM allocation
• Latency: ~38ms local inference on consumer ARM hardware
• Localization: Fine-tuned on 32 African languages, dialects, and regional policy/health contexts
• Deployment: Zero cloud dependencies, zero external network calls

Quickstart with llama-cpp-python

pip install llama-cpp-python huggingface_hub