reddit/r/LocalLLaMA
I benchmarked 13 models at 65K-128K context to find out what actually matters for agentic workloads
User Safety: safe