Your question is Quantization and Hardware Allocation Tradeoffs. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
Tell me how you would decide between different model quantization techniques and hardware allocation strategies given an inference cost constraint.
Assume you are supporting a production Fireworks AI deployment with a fixed monthly inference budget and a p99 latency target. Walk through how you would compare options, measure quality impact, and decide what runs on CPU versus GPU versus lower-precision accelerators. Focus on the decision framework, tradeoffs, and how you would validate the choice before rollout.