The first benchmark for enterprise AI solutions.

We test complete products, not just the models inside them, so you know what works before you buy. Every result below is observed, not claimed.

Read the results

Web search APIs tested: 7
SimpleQA questions each: 1,000
Observed accuracy spread: 75.9–91.1%

Vendor demos show what a product can do. Benchology shows what it actually does. We test competing enterprise AI solutions on the same real-world workflows and publish the evidence behind every result.

Our benchmarks

Two published evaluations. Each one tests a category end to end, on measured performance and on what the vendor will commit to.

1,000 search tasks

Benchology Benchmark 01

Web Search Tools

Retrieval accuracy (URL-exact match) across the same 1,000 search tasks.

Higher accuracy is better

Accuracy

70%75%80%85%90%95%

91.1%

90.0%

89.1%

84.8%

84.4%

83.4%

75.9%

Exa

Brave

You.com

Linkup

Serper

Tavily

Keenable

Search provider

● Same tasks● Same inputs● Observed resultsBar color = provider brand

Benchology Benchmark 01 · Reliability

What each vendor commits to

Scored disclosures, not measurements: what each vendor publicly commits to, 0–4 per axis. A low score means thin disclosure, not poor practice. Verified 2026-09-03, valid until 2026-12-02.

Exa

Tavily

Linkupfounded 2024

You.com

Brave Search API

Keenablepublic 2 weeks

All 30 scores as a table

Reliability disclosure scores, 0–4 per axis · verified 2026-09-03 · a low score means thin disclosure, not poor practice
Vendor	Price	Independence	Docs	Compliance	Procurement
Exa	2	4	2	3	3
Tavily	4	2	1	4	4
Linkup	3	3	2	3	3
You.com	4	1	1	2	2
Brave Search API	1	4	1	1	2
Keenable	insufficient history	3	0	0	1

The five dimensions

Each scored 0 at the centre to 4 at the outer ring.

Price: How much a buyer's cost basis has moved, and how existing customers were treated when it did.
Independence: Whether the vendor can be cut off or repriced by an upstream it does not control.
Docs: Whether the vendor has made the operational commitments a production integration depends on.
Compliance: Whether the vendor clears the first filter of a standard vendor security review.
Procurement: Whether an organisation with a procurement function can actually buy this, and how much friction it will meet.

See full results and evidenceData file · CC BY 4.0

A dashed axis means insufficient history, which is not the same as a score of zero — Keenable had been public two weeks at the verification date. Serper was scored and removed after scoring at the floor on all five dimensions; it stays in the measurements above and in the published data file.

Benchology Benchmark 01 — seven web search APIs over 1,000 SimpleQA questions each, September 2026. Accuracy is exact-source match against the reference URL; latency is mean response time; price is the published rate per 1,000 searches.
Provider	Accuracy	Mean latency	Price per 1,000
Exa	91.1%	1.71s	$7.00
Brave	90.0%	532ms	$5.00
You.com	89.1%	672ms	$5.00
Linkup	84.8%	2.21s	$5.00
Serper	84.4%	1.93s	$1.00
Tavily	83.4%	2.23s	$7.50
Keenable	75.9%	404ms	$4.00

Benchology Benchmark 02

Apollo vs Clay: how different are the people they find?

5 ICPs · 2,406 distinct people · 2,935 returned rows · measured 2026-09-12

Apollo only
Both (overlap)
Clay only

Bar length = people found by either vendor

ICPstress axis

Overlappeople found by either vendor · total 2,406

Valid row rateshare of returned rows that are valid

RevOps · San Francisco201–1000 employeesUS tech
52112144
Total 308
84.1%Apollo
92.6%Clay
Sales · Munich201–1000 employeesNon-English
113119407
Total 639
75.0%Apollo
88.2%Clay
Ops · Ohiomanufacturing, 201–500Industrial
161133327
Total 621
79.9%Apollo
83.0%Clay
Marketing · Singapore201–1000 employeesAPAC
5072245
Total 367
87.7%Apollo
79.8%Clay
Founders · AustinSaaS, 11–20 employeesSmall / SaaS
15184236
Total 471
84.7%Apollo
72.2%Clay

† Set 5's two encodings are not equivalent. Overlap figures are floors and unique-contribution figures are ceilings: the identity matcher recovers about 91% of true cross-vendor pairs, so the error runs one way.

See the full comparison

Example evaluation · GTM

See how Benchology evaluates a real workflow.

Here is how Benchology evaluates a GTM workflow, from understanding the work to testing qualified products.

01 · Define the workflow

First, we map how the work gets done.

Observed by Benchology

We define the steps and the success criteria. Every product runs the same workflow under the same conditions.

RevOpsDefines the system

SDRRuns prospecting

Account executiveOwns the account

MarketingShapes the signal

Define ICPSalesforce
Find target accountsApollo
Identify decision makersLinkedIn Sales Navigator
Research buying signalsSlack
Send personalized outreachGmail
Update CRMSalesforce

Benchologymaps the handoffs, systems, inputs, and success criteria before searching the market

Hands-on evaluation

Benchology enters the products.

ClaySame task running

Find US fintech companies with 500+ employees that recently hired a VP of Sales.

Company	Domain	Employees	Source
Brex	brex.com	1,200	Crunchbase
Ramp	ramp.com	1,000	LinkedIn
Galileo	galileo-ft.com	650	BuiltWith
Mercury	mercury.com	800	Crunchbase
Unit	unit.co	540	LinkedIn

ApolloSame task running

Find US fintech companies with 500+ employees that recently hired a VP of Sales.

Company	Domain	Employees	Source
Brex	brex.com	1,200	Crunchbase
Ramp	ramp.com	1,000	LinkedIn
Galileo	galileo-ft.com	650	News
Mercury	mercury.com	800	Crunchbase
Unit	unit.co	540	LinkedIn

ZoomInfoSame task running

Find US fintech companies with 500+ employees that recently hired a VP of Sales.

Company	Domain	Employees	Source
Brex	brex.com	1,200	ZoomInfo
Ramp	ramp.com	1,000	ZoomInfo
Galileo	galileo-ft.com	650	ZoomInfo
Mercury	mercury.com	800	ZoomInfo
Unit	unit.co	540	ZoomInfo

The same task, inputs, and success criteria are used in every product.

Market mapping

The products that claim this work.

Your requirementFind ICP accounts, decision makers, and verified buying signals.

Data

Apollo

ZoomInfo

Clay

Intent

6sense

Bombora

Common Room

Research

LinkedIn Sales Navigator

Perplexity

Outreach

Outreach

Salesloft

Open benchmarks

The results are already public.

Repeatable evaluations for the workflows enterprises actually run.

Private evaluation

Run the same test on your workflow.

Your requirements, your shortlist, the evidence, and a path to deployment.

Schedule a demo

Get one free evaluation.

Pick a time that suits you. Bring the workflow you want AI to improve, and we'll run one evaluation and show you the evidence behind it.

Schedule a demo