Research2d ago
Predicting the 2026 Open Model Gap
Nathan Lambert predicts that open-source models will narrow the performance gap with proprietary systems by mid-2026. He argues that synthetic data and refined distillation techniques will drive this convergence. This trend reduces the moat for closed-source labs. Developers can expect high-tier capabilities to shift toward open weights, lowering the cost of deployment.