• Sources: primary, discussion
  • Summary: NVIDIA released NeMo Switchyard as an open source library that routes each step of an agent workflow to a model chosen from a developer's own mix of open, proprietary, and NVIDIA models, selected on quality, latency, and cost. NVIDIA reports that LangChain routed 7 percent of calls to a larger model at 74 percent lower cost. The cost and quality figures in the announcement are reported by NVIDIA and its partners, and none were independently reproduced for this digest.
  • Why it matters: Per-step model selection across a mixed model set is being packaged as a library rather than as per-team glue.
  • Follow-up: Whether an independent party reproduces the cost figures.

send feedback on this story