[xhkkffbf]

This is great work. Has anyone used it enough to compare the lambda costs with the cost of running a comparable model on, say, OpenAI?