If one hour of AI is costing me 0.12€ is paying for frontier a cheaper option?

A user is evaluating the cost-effectiveness of running AI models like Qwen flash or Qwen 27b dense locally versus subscribing to a frontier model service. Their current setup, including a 7900xtx and 9800x3D, incurs an hourly AI cost of 0.12€, with electricity prices ranging from 0.11€ to 0.35€ per kWh. They are questioning if a frontier model, particularly if Deepseek flash is faster per 1 million tokens, would be a more economical option compared to their current self-hosted solution.