YoSaNoo/Forge-1-Little-Gemma-4-E2B

🚀 YoSaNoo / Forge-1-Little-Gemma-4-E2B

Hey there! 👋 Meet Forge-1-Little, a 5B parameter model that went to gym and came back WAY smarter than the stock Google's Gemma-4-E2B.

While big tech corps spend millions on massive parameter models, I trained this little beast on my laptop for 3 hours using pure CPU power and 16GB of RAM. And guess what? Data optimization > raw model size. It's already crossed 100+ downloads! 🎉

🧠 What makes it special?

• —Code Wizardry: It actually understands how to code. Toss a prompt at it, and it can write a functional Python + Pygame platformer with proper physics on the first try.
• —Big Brain Logic: Heavily tuned for agent environments, deep reasoning, and complex logic.
• —Polyglot: Don't worry, it didn't forget how to speak like a human. I mixed in a balanced multilingual dataset (40MB per language) so it stays friendly and articulate.
• —Laptop Friendly: Runs flawlessly on local hardware using GGUF quants. No high-end GPU required!

💻 Run it Locally (GGUF)

Via Ollama:

ollama run hf.co/YoSaNoo/Forge-1-Little-Gemma-4-E2B:Q4_K_M

Via llama.cpp:

llama cli -hf YoSaNoo/Forge-1-Little-Gemma-4-E2B:Q4_K_M

Or just drop it into LM Studio / Uncloth / Jan and start chatting!

Built by a solo dev.