Announcement
CoreWeave RL Rollouts is said to load new weights about 15x faster than a redeploy cycle
CoreWeave says RL Rollouts updates live deployments during reinforcement-learning training without touching in-flight requests.
TLDR
CoreWeave says its RL Rollouts load new weights into live deployments without touching in-flight requests, about 15x faster than a redeploy cycle. The company says Nvidia and You.com used the system to post-train Nemotron 3.5 Lightning for web search, raising BrowseComp accuracy from 36.97% to 45.45% while cutting tool calls by 30.24%.
Combined views
3.1K
1 Source, first seen ago
