Upgrading how?
Without knowing some salient details of how you're using the F4, and what's limiting you currently it's hard to advise.
The 144-pin foot-prints have allowed for an easier traversal across part families. There are also parts that can use interchangeable foot prints, allowing for increased sourcing opportunities.
For a better MCU and math, the Cortex-M7's can be a good path. The architected caching allows for effective use of external NOR Flash and unlocking for the effective code space.
As for "cost" the way to get there is the smallest die, and wafers yielding thousands of parts. Small die on newer tighter geometries is what decimates pricing. In respins for cost reduction, determine actual needs from your design related to code space, RAM and execution speed. Find the new parts that fit those requirements, find smallest package you can work with at a PCB level, or can reduce your PCB size or cost. Can you use a small QFN ? Can you halve your board size? Can you reduce the layers, or vias? Can your PCB accommodate multiple nested foot prints?
