Sign in Book a demo

How Otto compares on 1,000 searches

Each result is checked by a model against every requirement in the brief, then plotted against the real cost per task.

  • More verified matches per brief than any published result, on every chart below.
  • Under a sixth of the cost. Otto Medium finds 18.1 companies per brief for $0.09; the best published result finds 16.8 for $0.60. First on passed-per-dollar in every FindAll tier we ran.
  • 91.2% of delivered contacts verified on live people tasks, and 20 of 20 delivered in live lead verification.
07132026$0.00$0.57$1.15$1.72$2.30Cost per taskPassed per taskLowMediumHighOtto Find LeadsParallel Task UltraPerplexity Agent ProOpus 4.8GPT 5.5
  • Otto, at its Low, Medium and High effort tiers
  • Parallel Task Ultra
  • Perplexity Agent Pro
  • Reference models
Company-FindAll: Passed entities per task against cost per task. Higher and further left is better. Otto in ink; each published third-party result in its own colour, reference models as rings.
ProviderBest passed per taskAt cost per task
Otto Find Leads · High23.6$0.34
Parallel Task Ultra16.8$0.60
Perplexity Agent Pro11.5$0.06
Opus 4.84.4$0.13
GPT 5.518.7$2.10

On the public search benchmarks

Otto’s search on the standard benchmarks for search agents, October 2026. These measure accuracy and the FindAll score, not passed entities against cost, so they sit in their own table.

BenchmarkOttoMeasure
BrowseComp91.8%Accuracy
DeepSearchQA (DSQA)88.2%Accuracy
WideSearch56.1%Accuracy
Company FindAll15.4Score
People FindAll7.6Score
Cost per agentic search task$0.07Dollars
How we measured it

The task. Each provider receives the same natural-language brief and returns a list of entities. An entity passes when it verifiably matches every constraint in the brief at the time of the run. Two examples from the set: find heads of insights at mid-size FMCG brands in the UK, where a pass is a person who currently holds an insights leadership title at a UK FMCG company of 200 to 5,000 employees; and find audiobook publishers investing in AI voice, where a pass is a company that publishes audiobooks and has verifiable AI-voice activity in the last 12 months.

The axes. Y is passed entities per task, or the benchmark’s own score. Higher is better. X is the measured cost per task in dollars. Left is better. Otto is shown at its Low, Medium and High effort tiers.

The run. Otto figures come from our Find Leads benchmark run of 1,000 semantic searches, graded by a model against the pass criteria above. Competitor figures are published third-party benchmark results (2026); they are not our runs, and where a provider publishes its own result the higher number is reported.

The public benchmarks. BrowseComp, DeepSearchQA and WideSearch are graded for accuracy and FindAll by its own score, each by the benchmark’s published method; the cost is the average per agentic search task.

Otto and Grok Bot

Grok Bot is xAI’s general AI teammate. Its Sales Outbound role researches accounts, scores contacts and drafts email and LinkedIn messages in your voice, then sends the drafts you approve. The starting prompt in its own docs ends with a review list:

“Research the 25 accounts in this CRM view. Score them against our ideal customer profile (ICP) and recent intent, identify up to three relevant contacts per account, and draft email and LinkedIn outreach in the style examples attached. Skip anyone already in an active sequence. Return a review list; do not send or enroll anyone.”

GROK BOT DOCS · USE CASES
  • Sends from your own LinkedInConnection requests, messages, voice notes, video messages and InMail, from each teammate’s seat.
  • Runs the sequenceEmail, LinkedIn, WhatsApp, Instagram and calls on one canvas. A reply stops it.
  • Knows when to reach outSignal agents watch your accounts every day and tell you who is ready, with the reason.
  • Brings the replies homeEvery channel in one inbox, with Otto’s draft of the answer.
For outboundOttoGrok Bot, from xAI’s docs
Research accounts and score themYes. Every account scored against your ideal customer profile, with the reasons.Yes. Researches accounts and scores contacts with intent.
Find the peopleFind People from one sentence, on our database of 1B+ people, each match checked against every requirement in the brief.Identifies up to three contacts per account in its own example.
Draft email and LinkedIn messagesYes, in your voice, from the latest signal on the lead.Yes, in your voice.
Send from your own LinkedIn seatYes. Connection requests, messages, voice notes, video messages and InMail, from each teammate’s seat.No LinkedIn connector in xAI’s published list. Working LinkedIn through a browser agent instead puts your account at risk.
One sequence across channelsEmail, LinkedIn, WhatsApp, Instagram and calls on one canvas. A reply stops the sequence.Sends a draft once you approve it. No multi-step sequences.
Know when an account is ready41 ready signal agents: funding, new leaders, hiring, job changes, competitor engagement and more.Routines you set up from a prompt, such as a nightly research run.
Replies in one inboxEvery channel in one inbox, with the lead’s history beside the thread.Reads the mailboxes you connect.
CallsA dialer on every lead, with recording, transcript and a written summary.No dialer.
Your CRMTwo-way sync with HubSpot, Salesforce and Attio, and a push to Notion.A Salesforce connector, and others from its catalogue.
You in the loopApprove each one, or let it run.Keeps external actions behind your approval.

Grok Bot details from xAI’s own pages, read on 21 Oct 2026: Grok Bot use cases, use cases in the docs, how Grok Bot works and connectors. We have not benchmarked Grok Bot, so it has no numbers here.

Bring your hardest search

Half an hour. We run it live and you watch the pass rate.

Book a demo Start with a plan
The film