A New Hope for SLMs

Small language models (SLMs) have now reached a capability threshold where they can perform valuable work once exclusive to large frontier models. This development signifies a major shift in the AI landscape, as open-source models consistently replicate the abilities of closed-source leaders within a 6-12 month lag. Consequently, the future will likely see the vast majority of tasks handled by cheaper and more private open-source models, while frontier models tackle only the most complex scientific challenges. This trend shifts the focus from building the largest model to making capable SLMs reliable and production-ready for the 95% of use cases that don't require a frontier system.