Happy:2b LLM

Period

2026 – present

Client

Happy Center

An R&D project building a 2-billion-parameter language model on Google Colab Pro+, to explore how large language models work and dig into the attention mechanism. The first 500-million-parameter version correctly answered "Ankara" to "What is the capital of Türkiye?". The project is ongoing and keeps me close to the latest developments in LLMs.

Technologies