There isn't much about pricing, but this fragment suggests it will be economical mostly for light use cases. >“Currently, customers are paying for a lot of idl

There isn't much about pricing, but this fragment suggests it will be economical mostly for light use cases. >“Currently, customers are paying for a lot of idle compute in the form of virtual machines and GPUs that go unused,” I'm definitely looking forward to having a lot more competition in the "pay as you go LLM AI" space. Especially services that use models one can download and run on your own hardware once a good use case has been developed.