Why there is a lack of new 100B-120B models?

The article discusses the stagnation in the release of new models in the 100B-120B parameter range, highlighting that the most recent models, including GPT-OSS-120B and GLM-4.5-Air, are now at least three months old. It notes that newer releases have shifted towards smaller (25B-35B) or larger (200B+) models, raising questions about the future of the ~120B mixture of experts (MoE) family and whether new models will emerge by the second half of 2026. This trend is significant for AI practitioners as it may impact the availability of high-capacity models for specific applications, potentially limiting options for scaling and performance optimization.