Created a very simple benchmark to measure the actual knowledge cutoff of models.
It's surprising (or maybe not?) that OpenAI and Anthropic are the only big labs keeping their models fresh. Everyone else - including the chinese labs - is at least 12 months behind