Close Menu
  • AI
  • Content Creation
  • Tech
  • Robotics
AI-trends.todayAI-trends.today
  • AI
  • Content Creation
  • Tech
  • Robotics
Trending
  • OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI
  • A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels
  • Collectively AI Discloses Immediate Leak After Coding Agent Uncovered Credential – Unite.AI
  • Dragos Finds Immediate Leaks and Stolen Keys in AI Provide Chains
  • Lean and Mathlib Communities Certify Claude-Assisted Riemann Proof – Unite.AI
  • AWS Routes HyperPod Inference Throughout Areas From One Endpoint – Unite.AI
  • DailyAI Briefing: A Lean-checked Riemann proof, plus 4 extra
  • Anthropic Shares Drop 8.5% in Second Session Throughout AI Selloff – Unite.AI
AI-trends.todayAI-trends.today
Home»Tech»Exa 3.0 Search API Tops Perplexity, Courageous and Parallel on 8 Benchmarks

Exa 3.0 Search API Tops Perplexity, Courageous and Parallel on 8 Benchmarks

Tech By Gavin Wallace20/10/20265 Mins Read
Facebook Twitter LinkedIn Email
Microsoft Releases NLWeb: An Open Project that Allows Developers to
Microsoft Releases NLWeb: An Open Project that Allows Developers to
Share
Facebook Twitter LinkedIn Email

Exa has launched Exa 3.0, an improve to its net search API for AI brokers. The Exa 3.0 search API ships a bigger embedding mannequin and a less expensive search kind, Exa Lite. Dynamic Highlights is now typically obtainable. Exa claims that it leads Perplexity, Courageous and Parallel on 8 benchmarks. These benchmarks have been run by Exa, so you may deal with them as self-reported. For builders, the sensible change is price: Lite prices $4 per 1,000 requests.

TL;DR

  • Dimension: Exa tracks over 2 trillion URLs and serves 100 billion pages. Embedding mannequin measurement: Not disclosed.
  • Runs on: Hosted API solely. Exa lists configurable search latency of 450ms to 1s. No self-hosted choice.
  • Efficiency: Exa 3.0 Auto ranks first on all 8 Exa-run benchmarks towards Perplexity, Courageous and Parallel.
  • Greatest: 91.8% on BrowseComp, plus the widest lead on ATLAS (42.0% vs 37.2% for Perplexity).
  • Worst: 42.0% on ATLAS is its lowest proportion rating. Its BrowseComp lead over Perplexity is simply 0.3 factors.

What’s Exa 3.0?

Exa 3.0 is a search API launch from Exa Labs that upgrades retrieval high quality throughout each search kind. It’s obtainable via the API. Exa runs its personal crawler and an unbiased index. It isn’t a wrapper on Google or Bing outcomes.

How does Exa 3.0 enhance retrieval?

Exa attributes the positive factors to three modifications in its retrieval stack:

  • A bigger embedding mannequin: Exa skilled it with extra compute and knowledge than any earlier mannequin. Parameter rely shouldn’t be disclosed.
  • Dynamic Highlights, now typically obtainable: The mannequin selects tokens throughout all outcomes, not web page by web page. It provides extra tokens to higher pages and drops repeated data.
  • Billions of latest paperwork: New protection contains merchandise, native companies, flights, occasions, sports activities and climate.

In line with the Exa’s put up, Dynamic Highlights makes use of 95% fewer tokens than full page text whereas enhancing reply high quality. Highlights carry no further cost. That issues for brokers, the place context tokens usually price greater than search calls.

What’s Exa Lite?

Exa Lite is a brand new search kind, set with kind: "lite" on the /search endpoint, per the changelog. It runs on the identical index as Exa Auto. Auto spends extra compute on candidate era, highlights and rating. Lite helps question, end result rely, highlights, textual content, classes, area and date filters, location and moderation. Unsupported parameters return a 400 error naming the sector.

Exa now recommends 3 search sorts: lite, auto and deep. The quick, immediate and deep-reasoning sorts are legacy however nonetheless work.

How does Exa 3.0 carry out on benchmarks?

Exa in contrast every supplier’s highest-scoring single-step endpoint: Exa 3.0 Auto, Perplexity Net, Courageous LLM Context and Parallel Superior. All ran in a single agentic harness. The agent was GPT-6 Luna, with as much as 25 search rounds and 10 outcomes per search. Analysis occurred in October 2026.

#exa3-mtp-wrap iframe{show:block;width:100%!vital;border:0!vital;background:clear!vital}#exa3-mtp-wrap hr,#exa3-mtp-wrap p:empty,#exa3-mtp-wrap del,#exa3-mtp-wrap s{show:none!vital}

(operate(){var f=doc.getElementById(“exa3-mtp-frame”);window.addEventListener(“message”,operate(e){if(e.knowledge&&e.knowledge.kind===”exa3-mtp-height”&&e.supply===f.contentWindow){f.type.top=e.knowledge.h+”px”}})})();

The margins fluctuate broadly. ATLAS and WebCode present the clearest leads, at 4.8 and 5.6 factors over Perplexity. BrowseComp, DSQA and WideSearch leads are every below 1.5 factors over the runner-up. ATLAS is a benchmark Exa published itself on October 8, 2026.

On price, Exa plots agentic search rating towards price per activity. Lite scores about 79 at about $0.045 per activity. Auto scores about 80.5 at about $0.065. Exa says the pair units a brand new Pareto frontier amongst examined suppliers.

How does Exa 3.0 evaluate with Perplexity, Courageous and Parallel?

Function Exa 3.0 Auto Perplexity Search API Brave Search API (LLM Context) Parallel Search API (Advanced)
Listing value per 1K requests $7 (Lite $4) $5 $5 $5
BrowseComp (accuracy) 91.8% 91.5% 83.5% 63.1%
DSQA (F1) 88.2% 87.6% 78.6% 82.8%
WideSearch (F1) 56.1% 54.0% 54.8% 53.7%
ATLAS 42.0% 37.2% 28.2% 24.8%
WebCode 75.5% 69.9% 68.0% 64.5%
FinSearchComp 68.0% 63.4% 57.6% 66.0%
Firm FindAll (imply entities) 15.4 14.2 13.3 9.7
Folks FindAll (imply entities) 7.6 6.8 3.8 3.8
Embedding mannequin measurement Not disclosed Not disclosed Not disclosed Not disclosed

All benchmark scores come from Exa’s Exa 3.0 post and have been run by Exa. No vendor has printed matching outcomes. Costs come from every vendor’s personal pricing web page.

What knowledge does the expanded index add?

Exa states that it prioritized private AI assistant use instances. The official technical web page lists 1.27B ecommerce listings with costs and availability, and 211M+ local businesses. It additionally provides real-time information, sports activities, climate and market knowledge, together with flights, occasions, transit and podcasts.

The Exa Library buildings this knowledge into searchable collections. These embrace 320M individuals pages, 33M firm pages and 500M analysis paper pages. It additionally lists 19M SEC filings, 10M patents and 9M court docket opinions.

The place does Exa 3.0 fall quick?

  • Self-run evals: Exa picked the harness, endpoints and agent mannequin. ATLAS is Exa’s personal benchmark.
  • Skinny margins: A number of leads are below 2 factors, which can not survive different harnesses.
  • Lite limits: Lite rejects unsupported parameters with a 400, so some pipelines want Auto.
  • Increased Auto listing value: Auto prices $7 per 1,000 requests versus $5 for the three rivals.
  • Opaque mannequin particulars: Embedding mannequin measurement and structure will not be disclosed.

Key Takeaways

  • Exa 3.0 leads 8 benchmarks, all run by Exa itself.
  • Exa Lite prices $4 per 1,000 requests on the identical index as Auto.
  • Dynamic Highlights makes use of 95% fewer tokens than full pages, at no further price.
  • Greatest leads: WebCode (+5.6) and ATLAS (+4.8) over Perplexity.
  • Check by yourself queries earlier than switching suppliers.


Take a look at the Exa 3.0 announcement, the Exa Library, the changelog and the API docs. All credit score goes to the researcher of this mission. Additionally, be at liberty to comply with us on Twitter and don’t overlook to hitch our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

The put up Exa 3.0 Search API Tops Perplexity, Brave and Parallel on 8 Benchmarks appeared first on MarkTechPost.

api ar Benchmark search x
Share. Facebook Twitter LinkedIn Email
Avatar
Gavin Wallace

Related Posts

A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels

22/11/2026

OpenAI Ships GPT-Stay-1.1 for Voice Brokers, Cuts Idle Value to $0.02/Min

21/11/2026

Meet Pondering Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac

21/11/2026

Baseten Releases Bison-1 and Groundhog-1, Topping Choice Mannequin Index

21/11/2026
Top News

AlphaFold has changed the way science is done. Even after 5 years, it’s still evolving

The AI tool will tell you to stop slacking off

OpenAI Is Pissing Off a Bunch of Mathematicians—Once more

AI blog on Cybersecurity and Learning LLMs

China’s OpenClaw Boom is a Gold Rush For AI Companies

Load More
AI-Trends.Today

Your daily source of AI news and trends. Stay up to date with everything AI and automation!

X (Twitter) Instagram
Top Insights

Nvidia Connects Home Computers Into One AI Inference Cluster With PAIR – Unite.AI

04/09/2026

Physical Intelligence Team Reveals MEM for robots: Multi-scale Memory System Gives Gemma 3-B 4-B VLAs a 15-Minute Context For Complex Tasks

04/03/2026
Latest News

OpenRouter Cuts Three Suppliers After AI Immediate Publicity Findings – Unite.AI

22/11/2026

A Group of Researchers from Columbia and CertiK Introduce TensorLean for Formal Verification of GPU Kernels

22/11/2026
X (Twitter) Instagram
  • Privacy Policy
  • Contact Us
  • Terms and Conditions
© 2026 AI-Trends.Today

Type above and press Enter to search. Press Esc to cancel.