Feature/Bug Requests

Parasail should document what parameters they support and currently not support, like logit bias, XTC, DRY, etc.

The usage base pricing is great, but I would like a simple way for me to limit my spending, maybe just a simple input of number so can have a peace of mind that I won't burn my wallet without knowning it XD

You don't describe anywhere what you are logging, even though it was said that you log just the "amount" of input and output tokens, what is your policy?

Currently downloading the model, which is used for inference, can be expensive. For example, downloading Llama 3.3 70B takes over an hour (still running, so it might be multiple hours even). If preparation steps such as this weren't billed, and only the actual compute was, that would be great.

When I start an dedicated 70B model, it shows at low cost that it chooses 4x h200s and says "insufficient quota" or something like that. The fix for that is that I have to manually stop the dedicated model and start it again so that it only uses 2x of them, even though I had low cost selected.

I understand that Parasail is currently hosting Behemoth for free for a limited time; however, as far as I know, Behemoth doesn't provide commercial licenses, which means sooner or later either Parasail has to acquire a proper license from Mistral, or Parasail has to take it down, or we need to run it ourselves. What I'm looking for is similar to RunPod, where we can run HuggingFace models easily. The problem is running models on RunPod is never easy and you can't just drop a model's HuggingFace URL in it and expect it to work even though their web UI claims to be so. Also their documentation is pretty much non-existent. I hope in the future Parasail can provide ways to run models easily and thorough documentation and guidance as to how to use Parasail to run models from HuggingFace.

I have no idea how billing works, I was just told everything would be free for a month+, can someone give me more details. If I want to get my own model, should I just reach out to you directly?