Research2h ago
New Open Models Push Pareto Frontier
Three new releases, Laguna S2.1, Inkling, and Kimi K3, demonstrate that open-weights models now compete closely with proprietary counterparts. These artifacts prove training capacity is proliferating beyond a few labs. Researchers can now access high-performance checkpoints without relying on closed APIs. This trend lowers the barrier for specialized fine-tuning and academic verification.