Cartesia Managed Agents scores 1115 Elo on Artificial Analysis's Speech Agent Arena

Artificial Analysis says Cartesia Managed Agents, Cartesia's fully managed platform for building and deploying voice agents, scores an Elo of 1115 (95% CI 1086–1144, 685 samples) on its Speech Agent Arena, 98 points above Cartesia's earlier voice agent product, Cartesia Line (1017). The confidence intervals do not overlap. The results are Artificial Analysis's own.

Results it reports:

• On tasks that need tool calls, Managed Agents scores an Elo of 1112.
• Task Success Rate, the share of eligible tool-calling conversations that end with the correct final task-completing tool call, is 91.7%, up from 77.5% for Cartesia Line.

In the arena, people hold blind, live voice conversations with two hidden systems on the same task and pick the one they prefer. Managed Agents is a cascaded system: Cartesia's Ink speech-to-text and Sonic text-to-speech models run around an LLM the developer chooses. Artificial Analysis tested it with GPT-5.5 and the Skylar and Daniel voices.

Artificial Analysis lists the price as $0.06 per minute of call time ($3.60 per hour), with LLM usage billed separately by model.