Collection · benchmark v0.1 · ranked by accuracy gain in points
Biggest improvements between model generations
Where a newer version of a model gained the most accuracy over its predecessor on the same frozen questions, by category.
No model has a second generation on this benchmark version yet, so there is no improvement to measure. This page fills in the first time a successor version is run.
Nothing in this collection yet
No model has a second generation on this benchmark version yet, so there is no improvement to measure. This page fills in the first time a successor version is run.
Computed from the latest official run of every model on the current benchmark version. Mastermind is reported per item, never as a ranking criterion. All collections →