📢 Introducing Nomi-1.0-3b: A 3B Model That Thinks Like a 7B
Hey everyone! 👋
We just released Nomi-1.0-3b, a specialized mid-range LLM built on Llama-3.2-3B that punches way above its weight class.
🌟 What makes Nomi special:
⚡ Speed meets quality: ~60+ tokens/sec with only 8GB VRAM (RTX 4060, Laptops)
📝 Formatting Master: Trained specifically for structured reports, markdown, and clean tables
🐍 Coding Pro: Fine-tuned on Magpie-Pro to write Python with proper error handling
🌍 Bilingual Excellence: Fluent in German & English
🔒 Local-first: Perfect for privacy-focused deployments
The Goal: A "bridge" model that feels as intelligent as larger models but runs at 3B speeds.
Try it now:
Ollama/LM Studio Ready - Download the GGUF version
Works perfectly with standard Llama-3.2 chat templates
Apache 2.0 licensed - free to use & modify
🔗 Model Card: LazyLoopStudio/Nomi-1.0-3b
This is the first model in our Nomi-Series, with more optimized variants coming. If you're building AI applications that need speed + quality on consumer hardware, give it a try!
Feedback, benchmarks, and suggestions are always welcome 🙌