DeepMind researcher tests Kimi 3.1 Pro: strong benchmarks, poor real-world run

DeepMind researcher Samuel Albanie tested Kimi 3.1 Pro on a medium-difficulty prompt, finding the model took about 2 hours of reasoning despite strong benchmark scores. He acknowledged the sample size was only one.

2026-10-25 ~ 2026-10-25 · 2 related posts

• DeepMind researcher's first Kimi 3.1 Pro try goes poorly, with caveats on sample size — SamuelAlbanie · 2026-10-25

1 near-duplicate retellings: SamuelAlbanie