hi i can run any model on nim api but minimax-m3 allways return Internal Server Error, appear working on playground but not working on OpenAI sdk usage
I’m running into a weird situation where this model both works and doesn’t work for me. It fails when I try using it in a specific browser AI extension, but without any error logs, I can’t pinpoint the issue. However, it runs normally in OpenCode.
look like strange, does your OpenCode have the reasoning content displayed ? i think that the issue is when we active the thinking mode
did you ever figure out what was causing this? running into it as well :(
still nope :'(
i maybe have identify whats is wrong, that api endpoint does not accept “stream” argument from OpenAI sdk format, maybe because of OpenAI sdk is not using correctly the header or something like that, i will keep investigate but i think most issue is arround the header or accept format answer.
i will keep investigate but if is that we have to rework our workflow for our agent based on axios like the documentation sayed … seem really bad for us but stay tuned maybe i’m wrong
finally i found what is going wrong , nvidia nim required max_tokens on each call if you don’t add it to the payload so minimax-m3 is returned Internal Server Error <3
PS: OpenAI SDK still working fine , all the issue is just missing “max_tokens” params
Appear that Minimax-m3 have begun to get internal error also with max-tokens in buddy does any one have also encounter that new issue ?
seeing it too, was failing in playground for me last night