[kristianp]

Good point. When I use it, the inference doesn't seem very fast compared to the big providers, esp Time to First (non-reasoning)Token.