Full threadml_hardware·Wow! How are you able to achieve the cost reductions? Is it different hardware, software optimizations, or both? Also does this suggest that OpenAI is charging 6x markups on their models... >:(View on HN