> Once locals get to Opus levels I find it hard to imagine it would ever be cost efficient vs hosted/cloud i.e. you should always be able to run faster and/or

> Once locals get to Opus levels I find it hard to imagine it would ever be cost efficient vs hosted/cloud i.e. you should always be able to run faster and/or better models remotely at a comparable price since its just way more efficient due to batching