DeepResearcher
modelThe first deep-research agent trained end-to-end with RL in real-world web environments — live search and noisy open web rather than a curated RAG corpus — from Pengfei Liu's pre-SII GAIR lab (780★). Exhibited emergent planning, cross-source verification, and honest "I can't find it" behavior; released as DeepResearcher-7b.
Model Details
Architecture DENSE
Parameters 7B