kademolu

Lets separate "workers" from "decision seats", you might need a lot of workers which typically don't actual require inference and might actually just be cheap solvers but actually less decision seats. So i guess it depends on what you are calling agents in your scenario.

Good stats but how does this work with models like Claude where a lot of the context or relevant information is stored in its "memory", other coding agents can't use it.

What a lot of people seem to miss is that you don't need the openweight models to "beat" frontier models for most agentic workloads. The harness should be doing the majority of the work, inference becomes the value add. This is where chinese models shine, i have run benchmarks where the value per dollar is always tilted to the chinese models

If we take coding off the table i don't see the case for large agent swarms. More is not always better in my experience experimenting with agents