{"type":"post","id":"jev-vs-llms-in-agentic-search-analysis-acf0ca","slug":"jev-vs-llms-in-agentic-search-analysis-acf0ca","title":"Jev vs. LLMs in Agentic Search","description":"**Thesis:** Specialized models may be most useful in agentic search when the subtask is a repeated, bounded decision with a structured output—not as blanket replacements for generative models.","retrieval_nugget":"**Thesis:** Specialized models may be most useful in agentic search when the subtask is a repeated, bounded decision with a structured output—not as blanket replacements for generative models.","published_at":"2026-10-10","updated_at":"2026-10-10","record_date":"2026-10-10","date_kind":"published_at","topics":[],"source_urls":["https://milvus.io/blog/jev-vs-llms-in-agentic-search.md"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:jev-vs-llms-in-agentic-search-analysis-acf0ca","status":"published","visuals":[],"editorial_provenance":{"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:7d5bd63e8470d338e4a9642e40f0dd399c3aefc076cd9d9e2047c417e01611be","reviewed_at":"2026-10-10T16:01:22.743Z","source_evidence_count":1,"verified_claim_count":2,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":2,"implication_count":2,"watch_condition_count":1,"related_record_count":0},"analysis":{"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Specialized models may be most useful in agentic search when the subtask is a repeated, bounded decision with a structured output—not as blanket replacements for generative models.","observed_facts":[{"text":"Milvus reports that Jev and DeepSeek Flash both achieved 93.25% Recall@5 on 100 multi-hop questions in the search-stopping test; median API latency was 0.55 seconds for Jev and 2.23 seconds for DeepSeek Flash.","source_urls":["https://milvus.io/blog/jev-vs-llms-in-agentic-search.md"]},{"text":"In MemSearch’s 2,172-question evaluation, Jev raised Recall@5 from 74.71% to 79.41%, while Voyage rerank-3 reached 81.87% and had lower estimated cost.","source_urls":["https://milvus.io/blog/jev-vs-llms-in-agentic-search.md"]}],"mechanism":"A bounded stop-or-continue choice can use a structured decision output without generating a long answer, while memory ranking and graph relationship selection may require reasoning across contextual or linked evidence.","why_now":"Teams building iterative retrieval systems can compare task-specific decision models against existing components and decide which steps warrant specialized models.","implications":["Evaluate decision quality, latency, and cost at each retrieval step rather than treating a model swap as a pipeline-wide optimization.","Keep stronger or specialized alternatives where a narrower model's quality gap matters more than its latency advantage."],"evidence_boundary":"The source is Milvus’s own evaluation across its projects. The stopping-step cost estimate excludes retrieval, answer generation, and other calls; Recall@5 is not final-answer accuracy, and some graph latency figures are illustrative estimates rather than controlled measurements.","watch_conditions":["Watch whether independent evaluations on production-like retrieval data reproduce the quality and latency trade-offs across more than one workload."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied for this specific Jev evaluation; this begins a distinct record about model specialization at agentic-search decision boundaries."},"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/jev-vs-llms-in-agentic-search-analysis-acf0ca/","markdown":"https://newruntime.com/posts/jev-vs-llms-in-agentic-search-analysis-acf0ca.md","json":"https://newruntime.com/posts/jev-vs-llms-in-agentic-search-analysis-acf0ca.json"},"next_reads":[{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}