Fetching from the wire…
01
02
03
04
05
06
07
08
Models2026-08-15 · source-backed
The specialized search agent decomposes queries into subqueries, gathers evidence and curates context. Published numbers: 70% answer correctness on OfficeQA Pro V2 at about $1.15 per task, 3.5x fewer tokens at equal performance on Harvey's LAB benchmark, 8-11 second median retrieval latency, $0.016-$0.023 per standard query. Mixedbread It ships standalone and as a subagent inside a frontier model. Retrieval as a purchased specialist rather than a RAG pipeline you maintain is the architectural bet worth watching.
Each link below shares sources, entities, or timing with this story.
Harvey uses Claude / Shared entity: RAG / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Harvey uses Claude); both cover RAG; overlapping topics (agent, context).
Harvey uses Claude
Linked by a graph relationship (Harvey uses Claude).
Linked by a graph relationship (Harvey uses Claude).
Harvey uses Claude / Shared entity: RAG / Shared topic / Earlier coverage
Linked by a graph relationship (Harvey uses Claude); both cover RAG; overlapping topics (agent, context, query).
Harvey uses Claude / Shared entity: RAG / Earlier coverage
Linked by a graph relationship (Harvey uses Claude); both cover RAG; earlier RAG coverage from 2026-07-08.
Linked by a graph relationship (Harvey uses Claude); both cover RAG; earlier RAG coverage from 2026-07-07.
Harvey raised Sequoia
Linked by a graph relationship (Harvey raised Sequoia).
Linked by a graph relationship (Harvey raised Sequoia).