ZoomInfo (NASDAQ: GTM) has released GTM Bench, a benchmark that evaluates large language models and AI agents on go-to-market tasks such as building prospect lists, enriching records, scoring accounts, and reaching decision-makers, according to a company press release.

Version 1 of the benchmark covers more than 20 job types, four systems, and three models. Results are graded on two axes: how much of the requested work the system completes, and how much of the returned data traces to a verifiable source. A confident but incorrect answer scores negative.

In the initial run, ZoomInfo's own GTM.AI product scored a GTM Bench Index of 77, compared with 47 for Apollo, 36 for Exa, and 31 for open-web search. ZoomInfo states its system completed 98% of a reference operator's work product and returned 478 verifiable records per 1,000, against 7 to 35 for the other systems tested. The cost per task was reported at $0.79.

ZoomInfo acknowledges the benchmark is vendor-run and that grounding is measured against its own verified records rather than an independent audit. The company also discloses four categories where its system holds little or no advantage, including pure copywriting and data stored in a customer's own CRM.

The benchmark is versioned and will be updated with major model releases. A second version is planned to add multi-step agentic workflows, international coverage, and an owned-data axis. Other vendors may submit their systems to be tested and added to the leaderboard.

ZoomInfo's GTM.AI product connects to its database of more than 100 million companies and 500 million contacts and integrates with platforms including Salesforce Agentforce, HubSpot Breeze, Microsoft Copilot, and Google Workspace.