STORY · PRODUKTER_
Open-source models beat GPT-5.6 Sol on cost-efficiency for retrieval
Castform and Neon demonstrate that a 4B open-source model can be trained via reinforcement learning to solve retrieval tasks as well as GPT-5.6 Sol, but at 100 times lower cost. Castform automates RL post-training of open-source models by using existing enterprise data as training material.
WHY IT MATTERS
This drastically reduces the cost of agentic AI and makes it possible for small teams to use local models instead of expensive API calls. It shows that frontier models are no longer necessary for all tasks.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.