Summary
Pinecone announced general availability of Nexus, a knowledge engine that compiles an enterprise's proprietary documents, databases, and workflows into a governed, agent-ready knowledge layer delivered to agents in a single call and queried through KnowQL, a declarative query language built for agents. On Sierra's open τ-Knowledge benchmark, an agent using Nexus posted the top score, and pairing GPT-5.5 with Nexus held accuracy roughly flat while cutting cost per task by 77%.
What changed
Nexus moved to general availability. It turns proprietary data and workflows into a structured, governed knowledge layer delivered to agents via KnowQL and can deploy in the customer's own cloud. On τ-Knowledge, GPT-5.5 with Nexus solved 47.4% of tasks (top score) versus 46.4% for GPT-5.5 alone at 77% lower cost per task; GPT-5.2 with Nexus reached 36.1% versus 32.2% unaided at 80% lower cost.
Why it matters
It reframes enterprise RAG as a governed knowledge layer rather than raw vector search, arguing that better knowledge, not a bigger frontier model, drives agent accuracy and cost. The benchmark framing positions Pinecone as infrastructure that lifts cheaper models to rival frontier setups at a fraction of the cost.
Evidence excerpt
On its debut on τ-Knowledge, Sierra's open benchmark for the most demanding enterprise knowledge tasks, an agent using Nexus as its knowledge layer posted the top score... GPT-5.5 with a Nexus knowledge layer solved 47.4% of tasks... while cutting cost per task by 77%. Agents query the layer through KnowQL, a declarative query language built specifically for agents.