Sounds like an incremental change with context size change, new "box" type, and composable images.
As it's minor change did not expect much, but having a higher context helps for sure. I will own the errors for passing bigger context.
Well, they do need to try to compete with SubQ’s longer context and image processing that has been coming out ahead in benchmarks, plus the thousand self-hosted options that now exist.
Well, they do need to try to compete with SubQ’s longer context and image processing that has been coming out ahead in benchmarks, plus the thousand self-hosted options that now exist.