Zoom Info launches GTM Bench to set industry benchmark for AI sales tools
Info Technologies Inc. The company recently announced the launch of the GTM Bench benchmark system, which is specifically designed to evaluate the performance of large language models and artificial intelligence agents in real marketing tasks. GTM Bench scores AI systems from two major dimensions: "Task Completion" measures the proportion of the system completing the required work, and "Data Traceability" measures the proportion of returned data that can be traced back to real, current sources. In the first round of testing, ZoomInfo's GTM.AI scored a comprehensive index of 77 points, while Apollo scored 47 points, Exa scored 36 points, and Open Web Search scored 31 points. The ZoomInfo system completed 98% of the operational work, returning 478 verifiable records per 1,000 records, far exceeding the 7 to 35 of other systems, and costing only $0.79 per task. It is worth mentioning that ZoomInfo also frankly announced four scenarios in which it performed poorly, including pure copywriting and its own CRM data that cannot be accessed by external tools. The measurement of data traceability is also based on ZoomInfo's own verification records, rather than being completed by independent audits. The second version of GTM Bench will add multi-step agent workflow, international coverage and own data evaluation dimensions. Other data providers and AI agent developers can also submit systems to participate in testing. As the contextual data layer of ZoomInfo, GTM.AI provides access to more than 100 million companies, 500 million contacts and billions of signal data. Salesforce Agentforce, HubSpot Breeze, Copilot, Claude, ChatGPT, etc. are integrated.