Glean Highlights Specialized Agentic Search Model to Cut Latency and Costs

According to a recent LinkedIn post from Glean, the company is highlighting “Waldo,” described as its first agentic search model designed to orchestrate how enterprise queries are decomposed, which tools are invoked, and when to hand off to larger frontier models. The post frames Waldo as a specialized layer that plans retrieval and context-building before a more expensive reasoning model generates the final answer.

The company’s LinkedIn post suggests that Waldo targets retrieval-heavy workloads where latency and cost are key constraints, positioning it as a complement rather than a replacement for frontier LLMs. Quantitative benchmarks in the post indicate that Waldo runs roughly 10x faster per LLM call, with about 250 ms P50 latency versus roughly 3 seconds for Glean’s default reasoning model.

In Glean’s internal harness, the post reports that these improvements translate into approximately 50% lower end-to-end latency and about 25% lower token cost for applicable use cases. For investors, these metrics imply potential margin benefits and improved scalability for Glean’s AI-powered enterprise search and productivity offerings, particularly for large customers with high query volumes.

The post also notes that Waldo is built on NVIDIA’s Nemotron 3 Nano and has been post-trained for search planning, with NVIDIA and Thinking Machines Lab referenced as partners. This alignment with NVIDIA’s model ecosystem may enhance Glean’s credibility in enterprise AI infrastructure and could support co-marketing or technical integration opportunities that strengthen its competitive position against other AI-native workplace search platforms.

Strategically, the post positions Waldo as an early example of specialized models tailored to high-demand, well-defined tasks where latency and cost matter, rather than relying solely on general-purpose frontier models. If this architecture gains adoption, Glean could benefit from differentiated performance characteristics, potentially improving customer retention and upsell prospects in a consolidating enterprise AI market.