• LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
• agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
• reranking — A second pass that re-scores retrieved candidates by reading each one against the query, fixing the ordering that fast vector search got approximately right.

In an agentic search pipeline every quality step, including rewriting, reranking, and snippet summarization, is itself a blocking LLM call, so decode speed sets how deep the pipeline can go. Per-step timings against Gemini 3.1 Flash Lite and Claude are published with a reproducible harness.

• articleRise Of Realtime Subagents
• articleUltra Fast Apply Edit With Mercury Coder
• articleMercury Refreshed
• articleIntroducing Mercury

• articleWeb Search BenchmarkOpenRouter editorial sitemap
• articleMercury 2 On PinchbenchInception editorial sitemap
• article➤ Leaderboard: ➤ Methodology: ➤ Stirrup, our open-source agent harness: Search AArtificialAnlys

Checking sign-in…

Loading comments…