Cartesia launches a complete voice-agent platform for enterprise calls
Managed Agents combines speech models, reasoning and telephony, and leads two independent benchmarks at launch.
Cartesia has launched Managed Agents, an end-to-end platform for companies building voice agents for customer service, scheduling, retail returns and internal help desks.
The product combines telephony, speech recognition and synthesis, reasoning, tool calls and knowledge bases in one system. Cartesia says it co-designed the voice models and agent harness so turn-taking, interruptions and backend actions can be optimized together. It supports 44 languages, while mid-call switching currently works across English, Spanish, Hindi, French and Japanese.
At launch, Managed Agents ranks first among 30 systems on Artificial Analysis’s Speech Agent Arena, with an Elo score of 1115 from blind human comparisons. It also leads ServiceNow’s EVA-Bench across more than 200 enterprise scenarios, scoring 87.4% for accuracy and 90.7% for experience. Those tests used GPT-5.5 as the reasoning model, so the result measures Cartesia’s voice layer and orchestration rather than a standalone language model.
Cartesia says the platform has already handled tens of millions of conversations in finance, retail, hospitality and healthcare. It also includes outbound call campaigns and integrations with business systems.
The launch pushes the voice-model company into direct competition with complete agent platforms such as ElevenLabs Agents. Benchmark leadership is useful evidence, but production reliability, price and deployment controls will matter more for enterprise buyers; Cartesia has not yet disclosed all of those details publicly.
