Nathan Lambert@natolambertOct 28, 2026, 18:21I feel a need to defend the AAv5 index and open models at the same time. Benchmarking is messier, more opinionated, and trickier than most people give credit to.打开原帖#511482