I guess so. 4.8 + 4.8 > Fable 5 is interesting, though not particularly game changing. (The others all fuse frontier models. Which is an argument for using those frontier models more. Not less.)
Yeah, all that's really saying is a weaker model with a better harness can beat a stronger model with a worse harness, specifically on the DRACO benchmark
This isn't really a surprising result. Needs more evidence to make a broader claim.
