Language Model News

Moonshot

All articles

2026/09/15

  1. Cognition's SWE-2 is a coding model built on Kimi K3

2026/09/08

  1. Kimi K3 holds a million tokens at one flat rate

2026/09/02

  1. What prompt caching actually costs, and where the tenth does not hold
  2. Long context: who reprices the whole request, and who does not