Skip to main content

You.com Launches 'Highlights' Web Search Mode, Claiming 95.17% on SimpleQA

2 SEPTEMBER 2026·2 MIN READ·1 SOURCE·Trusted source

You.com has introduced a new extraction mode for its Web Search API called Highlights, which achieved a 95.17% score on the SimpleQA benchmark.

You.com Launches 'Highlights' Web Search Mode, Claiming 95.17% on SimpleQA

Key takeaways · 3

  • 01

    Highlights achieved a 95.17% score on SimpleQA and 66.93% on RetrievalQA.

  • 02

    The feature increases cost per question by 11% compared to snippets, but lowers cost per correct answer by 5%.

  • 03

    Latency for You.com with Highlights was reported at 695ms p50.

How Highlights Works

You.com introduced a new extraction mode for its Web Search API called Highlights on September 1, 2026. [1] Setting the `extraction_mode` parameter to `highlights` replaces default snippets with an array containing passages that specifically address the query. [1] A model identifies text spans that answer the query and returns them verbatim, preserving formatting for numbers, dates, tables, and code blocks. [1]

Benchmark Performance

On the SimpleQA benchmark, You.com reported the Highlights feature reached a score of 95.17% with a 695ms p50 latency. [1] The feature also scored 66.93% on the RetrievalQA benchmark, though it scored slightly lower than full-page extraction on the FRAMES multi-hop reasoning test. [1] The new mode increases the cost per question by roughly 11% compared to standard snippets, but the cost per correct answer is about 5% lower. [1]

What it means

By offering targeted passage extraction rather than raw full-page returns or generic keyword snippets, You.com is optimizing for retrieval accuracy while controlling downstream LLM token costs. The reported 695ms latency provides a significant speed advantage compared to competitors like Exa and Parallel, which also cleared the 95% accuracy mark on SimpleQA but took substantially longer to return results. What the sources don't address: How the API performs on highly unstructured or dynamically rendered web pages that resist standard targeted passage extraction.

Providing LLMs with precise, verbatim passages rather than entire web pages or keyword fragments can improve retrieval-augmented generation accuracy and lower token consumption. The new extraction mode highlights a trend toward moving the retrieval-filtering burden off the generation model and into the search API itself.

Why it matters
Daily session

Turn this story into practical AI skill after launch.

Get the release link for daily sessions built around your role and industry.

Join the waitlist

How this developed

  1. 2 September 2026

    You.com Launches 'Highlights' Web Search Mode, Claiming 95.17% on SimpleQA

  2. 2 September 2026

    Event created from source cluster.

Sources

AI fluency, one session a day, built for your work.