DeepSeek-R1 shows 14.3% and 11.3% on Vectara—is it usable for summaries?

https://fire2020.org/when-models-disagree-what-contradictions-reveal-that-a-single-ai-would-miss/

Every time a new model drops, the industry treats the latest benchmark scores like Olympic gold medals. DeepSeek-R1 has been making waves, and the recent numbers hitting the Vectara HHEM leaderboard have triggered the usual frenzy