Jacob GPTProto

I run gptproto — an API layer that puts 200+ models behind one OpenAI-compatible endpoint, at 10-30% below official rates. Most of what I post here is about what inference actually costs: real numbers from our own usage rather than list prices. If I get something wrong, tell me in the comments and I'll fix it.