Continue benchmarking various AI models to replace #watchdog's primary model and see several "top tier" models show a high error rate, bad latency, poor moderation ability vs. open source models that are supposedly "less intelligent" - exactly why I decided to run my own benchmark instead of trusting artificialanalysis #watchdog