logoalt Hacker News

aleqs • today at 7:14 PM • 0 replies • view on HN

We know anthropic nerfs models all the time, openai probably does the same. There is no reason to believe the benchmarks/numbers they release are even measured with the same effective model that customers would get.