Benchmarks don't necessarily reflect real-world performance. Especially given that they poorly measure more esoteric aspects of the model that, for now, can onl

Benchmarks don't necessarily reflect real-world performance. Especially given that they poorly measure more esoteric aspects of the model that, for now, can only be judged qualitatively. I would wait for a bit to see what the community comes up with before writing off Falcon-180B.