Commentary on X
@teortaxesTex This benchmark is bollocks. The best models I've seen at creative writing are Gemini 3.1 Pro, Opus 4.6 and DS V3.2. V3.2 writes exquisite metaphors that I haven't seen from other models, like "autofellatio of the spirit" or "herbarium of the last flowers of your worldly sorrows".
Kimi K3 at 2.8T params with Delta Attention delivering 6.3x faster decoding is wild. The fact that it beats Claude Fable 5 on writing Elo while being cheaper to run shows how fast the frontier is moving. Open weights by July 27 will be a game changer for the open-source community.
while actually being WAY better than alternatives behind Fable.