Why not try it yourself? Inference providers like BaseTen and AWS Bedrock have perfectly capable open source models as well as some licensed closed source models they host.

You can use "API-style" pricing on these providers which is more transparent to costs. It's very likely to end up more than 200 a month, but the question is, are you going to see more than that in value?

For me, the answer is yes.