ruslandev/llama-3-8b-gpt-4o-ru1.0

ruslandev/llama-3-8b-gpt-4o-ru1.0 is an 8 billion parameter instruction-tuned Llama 3 model developed by ruslandev, fine-tuned for the Russian language. It leverages a high-quality dataset derived from GPT-4o, focusing 80% of its training on Russian examples. This model demonstrates strong performance in Russian language tasks, surpassing GPT-3.5-turbo and matching other multilingual models on the MT-Bench benchmark.