SID.ai unveils agentic search model SID-1: 1.9x recall, 24x faster

baseten · x · 2026-11-18

SID.ai (YC-backed) announced SID-1, an agentic search model claiming 1.9x better recall than embedding-only approaches, 24x faster, and beating frontier models including GPT-5 on complex search tasks, trained with 1k+ QPS RL rollouts. The lean SF/Zürich team is backed by Y Combinator, Canaan, Rebel, General Catalyst, with advisors from Anthropic, DeepMind, OpenAI, Cursor and more. The post also announces the chalk-talk lineup for Baseten's All-Gather event: Kyle Corbitt (CoreWeave), Max Rumpf (SID.ai), and Alex Nails (RadixArk).

Original post →

More from Companies & People

Companies & People channel →