Monitor spend and surface optimization opportunities across providers
Switch nemotron-super-49b to Together AI for non-latency-sensitive workloads
Enable request batching for Code Review Agent
Use Fireworks AI for llama-3.3-70b