Does this include token reduction and price routing?

[TomLucidor]

Coming from the other cmux-like tools, I wonder if things like these can have ways to make SLMs more powerful by using performence-aware routers. And if some tasks are hard, maybe context pruning and caveman/ponytail output prompts can come pre-wrapped