[v5] Cost Optimizations: With using specialized local models

[hemant12d]

After testing, currently the runs approximately cost around 7 to 9 cents per job application. Using local models can make it even cheaper.

[srmanda-cs]

Assigned it to you Hemant, it is already very cheap, let's see how much cheaper we can make it without losing quality.