Company overview
You.com built the Answer API to round out its Web Search API portfolio, delivering a synthesized, fact-checked answer with verified citations from a single query: 93.48% accuracy on SimpleQA, 2.67-second p50 latency, and $5 per 1,000 calls.
Baseten powers the inference behind that stack, giving You.com frontier-level answer quality without routing every query to a frontier model, at a fraction of the cost.
The Challenge
You.com gives agents and chatbots real-time, cited, hallucination-resistant answers by closing the gap between LLMs and the live web.
A synthesized answer endpoint had been on You.com's roadmap for a while. The economics were the constraint: routing every query through a frontier model puts the cost per call above the price You.com wanted to offer developers. Open-weight models reaching parity on bounded synthesis tasks is what made the Answer API viable, and it brought the team to the Baseten Inference Platform.
"We think open-weight models plus web search is a powerful alternative to routing everything through frontier model providers, at a fraction of the price and comparable quality for our developers." — Saahil Jain, CTO, You.com
Solution
Baseten powers the inference layer behind several stages of You.com's answer pipeline, running open-source models across dedicated, multi-region deployments.
Generation
You.com serves a range of open-source LLMs on Baseten, including newer models like GLM-5.2 and Kimi K3, chosen to sit on a Pareto-efficient frontier of accuracy and cost. These models handle the synthesis step behind the Answer API, turning retrieved passages into a cited, verified answer instead of a list of links.
You.com evaluates candidates on accuracy, p50 latency, and cost per call against its own benchmark set, and serves whichever open-weight models currently clear the bar. Synthesis is a bounded task: the model reads pre-filtered evidence and produces an answer where every claim traces back to a specific excerpt, with no reliance on world knowledge or open-ended instruction following. On that shape of workload, the accuracy gap to a comparable proprietary model came in under half a percentage point.
Key Baseten platform capabilities You.com relies on include:
Dedicated, multi-region deployments for low-latency serving close to users.
Zero Data Retention (ZDR) to meet enterprise compliance requirements.
Baseten Model APIs for frontier-grade open-source LLMs like GLM-5.2 and Kimi K3
Results
The Answer API is now running in production on Baseten, inside a stack serving tens of millions of queries a day. Compared with routing synthesis to a frontier model, You.com saw:
Under an hour to migrate
30% lower cost per call
Open-weight model accuracy on par with closed source models
Matching developer inference needs to a fast, cost-effective open-source model, backed by production-ready infrastructure, is what lets You.com deliver frontier-level answers at a fraction of what a frontier model would cost.
Get Started
You.com's Answer API is live today. If you're building an agent, chatbot, or research product that needs reliable answers from the live web, see the Answer API docs to get started, or read You.com's engineering write-up on why they chose open weights.
And if you're evaluating open-source models for your own inference stack, talk to us about what Baseten can do for your production workloads.
