Just a thought, I'm not sure if AH plans to host Deepseek v4? They're good LLMs (I’m using Pro), and there's a tough confusing world out there with bugged providers, which makes users have a worse experience with OH. 🤔 It could be nice to offer some hosting which works as expected. (example issue)

Do you mean provide it through the proxy? We have flash already but now that you mention it maybe we don’t have pro

Well, I was thinking about serving it. Would that be possible?

Possible but probably not cost effective unless we make it the default model

nod though sometimes they really are messing things up. Just now I’m looking at Deepseek Pro from

openhands/

(it is available), and it seems it’s not cached at all or almost at all. This could also be from the sdk (investigating), but it definitely happens that some providers seem cheap, but they don’t cache so… they’re the most expensive in reality

This is specifically why I hold OpenRouter in low regard. As somebody who seeks out lab diversity, they offer models that straight-up do not work. This isn't something that happened once. I found it repeatedly when trying to test underdog models. Not just them, tho. I've seen errors messages that were stack-traces thru 3 different providers, Generally speaking, I don't really expect caching to work unless I'm getting inference directly from the lab.

Ugh yes, I think OpenRouter has some user choice though, I seem to recall you can forbid certain providers