Two Minute Papers · 2026-07-29 · notable
Two Minute Papers: 'Kimi K3 Just Broke The Economics Of AI'
Károly Zsolnai-Fehér walks through Moonshot's Kimi K3 — the 2.8T open-weights MoE that has reset what a frontier-class model costs to train and serve — for a general engineering audience.
Two Minute Papers explains why Moonshot's 2.8T open-weights Kimi K3 shifts the cost floor for frontier-tier models.
What is it?
'Kimi K3 Just Broke The Economics Of AI' is Károly Zsolnai-Fehér's July 29 breakdown of Moonshot's Kimi K3 release — a 2.8T-parameter mixture-of-experts model that shipped open-weights on Hugging Face with 1M-token context, aimed at a general engineering audience rather than researchers.
How does it work?
The video frames Kimi K3 against the closed frontier — walking through the model's mixture-of-experts routing, its published training-compute footprint, and the open-weights release strategy — to explain why open labs can now match closed flagships on major benchmarks without a matching hardware bill.
Why does it matter?
Two Minute Papers is where many engineers and non-specialists first meet frontier model launches, and Kimi K3's economics claim reshapes the build-vs-buy math for teams that assumed only closed labs could ship at this scale. Making the cost side legible helps teams re-evaluate whether an open-weights deployment is now a serious option.
Who is it for?
Engineers, tech leads, and curious viewers deciding whether to bet on open-weights frontier models.