Comment by impure

I recently made an AI Agent and surprisingly coding with DeepSeek V4 Flash is quite cheap. It probably has to do with the aggressive prompt caching. I'm using OpenRouter with Novita AI as the preferred provider.