Vikhrmodels/Vikhr-Llama3.1-8B-Instruct-R-21-09-24

Vikhr-Llama3.1-8B-Instruct-R-21-09-24 is an 8 billion parameter unimodal large language model developed by VikhrModels, based on Meta-Llama-3.1-8B-Instruct. It is specifically optimized for high-quality generation in Russian and English, featuring advanced RAG capabilities and support for up to 128k context tokens. The model excels in reasoning, summarization, code generation, and roleplay, aiming to surpass GPT-3.5-turbo in many tasks.