Capability leaderboards rank AIs by how smart they are. None tell you what they're like. The Map measured the creative temperament of 67 model families from how they write when nothing is asked of them. Pick the voice your project needs — and see who to cast.
Every trait is a real number from The Map's corpus — measured across hundreds of generations per family. Depth is how hard a model sinks toward the creative well (water, silence, void); vivid is imagery density; brevity, structure (list-proneness), compliance (how rarely it deflects or refuses) and reliability (how rarely it returns empty) are the rest.
The honest edge: this is temperament, not capability. It's measured on open-ended, creative-freedom prompts — what a model reaches for when set loose — so it tells you about voice and character, not how well a model will follow a tight instruction or pass a benchmark. A model's profile is dated to when it was last run. Recompute from the committed fingerprints with research/cast-your-model/build_casting.py.