I was road testing tool calling in LM Studio a week ago against a few models marked with tool support, none worked, so I believe it may be a bug. Had much better luck with llama.cpp’s llama-server.