s/efficiency/performance/g

i still use sonnet 4.6

Better than Kimi k2.6 for the perf/cost ratio?

How do you use it? Cloud? Home? How privately do you use it?

The interesting question is whether this Pareto frontier holds across real-world workloads or just benchmarks. Efficiency claims need scrutiny — but if true, it shifts the cost calculus for running open models at scale.