Kumru LLM: Turkish 7.4B Parameter Large Language Model
Kumru is a 7.4B parameter decoder-only Turkish LLM pre-trained from scratch on 500GB of Turkish corpora, surpassing significantly larger multilingual models on the Cetvel benchmark.
LLM
VBART: The Turkish LLM
VBART are the first Turkish sequence-to-sequence LLMs pre-trained on a large corpus from scratch, surpassing prior state-of-the-art on summarization, paraphrasing, QA, and more.