Can a Local LLM Run My AI Assistant?

An AI developer tested two local large language models (LLMs) running on RTX 3090 GPUs against the Claude model for an AI assistant handling 27 real tasks. The smaller 30B model scored poorly with errors, while the larger 122B model matched Claude's performance at a fraction of the cost, enabling local deployment.