Comment by llmslave

These models are never as good, the benchmarks dont tell the full story