Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontier

Substack · Read

• Surveys a wave of open releases — Thinking Machines’ Inkling, Tencent’s Hy3, Poolside’s Laguna-S-2.1, DeepSeek’s V4-Flash update, and Moonshot’s Kimi K3 (the largest open release in some time).
• Core claim: “capacity to train strong models is proliferating” — the open-model frontier is no longer defined by one or two labs.