Open labs embrace continued post-training over new model generations

A Reddit discussion argues that open labs are shifting from releasing entirely new model generations (e.g., Qwen 1.5, 2, 2.5, 3, 3.5) to continued post-training on existing models, citing recent releases as evidence. The post suggests this approach is more effective.