Voice model company Cartesia launched Managed Agents on Oct 29, an end-to-end voice agent platform for enterprise phone calls that tunes its own speech-to-text and text-to-speech models together with the agent harness, and bundles telephony, knowledge bases, tool calls and outbound batches. Cartesia says it ranks #1 on Artificial Analysis's Speech Agent Arena with a 1115 Elo and scores 87.4% accuracy and 90.7% experience on ServiceNow's EVA-Bench (vendor-reported, with GPT-5.5 as the reasoning model). Launch pricing is 50% off at $0.03 per minute.
Key Takeaways
- ✓On Artificial Analysis's Speech Agent Arena blind preference test, Managed Agents ranks #1 of 30 agents with a 1115 Elo, ahead of Gemini 3.1 Flash Live Minimal (1099), Grok Voice 3.0 (1064) and GPT-Live-1 (1041) (Cartesia blog, as cited by the vendor)
- ✓First-attempt results on ServiceNow EVA-Bench across 200+ customer service, IT service and HR scenarios: 87.4% accuracy and 90.7% experience, versus 73.7%/71.2% for Grok Voice 3.0 and 59.0%/61.6% for GPT Realtime 2.1 (EVA-Bench)
- ✓All these benchmark runs use GPT-5.5 as the reasoning model; Cartesia's internal evaluation says 85% of calls had no turn-taking issues, versus 67% for other agents
- ✓Launch pricing is 50% off at $0.03 per minute; Cartesia says the platform has already handled tens of millions of conversations in finance, retail, hospitality and healthcare
- ✓It supports 44 languages; English, Spanish, Hindi, French and Japanese can switch mid-call today, with the rest coming soon
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background and pain points
Enterprise phone agents are usually built by chaining separate speech-to-text, LLM and text-to-speech services. That cascade is controllable and can call tools, but tuning each piece separately tends to produce talk-over, agents that keep speaking when interrupted, and awkward pauses. End-to-end speech-to-speech models sound more natural but are hard to control for complex workflows. Cartesia, known for its Sonic text-to-speech and Ink speech-to-text models, now offers the whole stack as a hosted platform, competing directly with ElevenLabs Agents and the realtime voice offerings from OpenAI, Google and xAI.
Architecture highlights
The launch blog by Sauhard Jain says Managed Agents tunes listening, turn-taking, reasoning, tool calling and speech as one system instead of assembling third-party parts; CEO Karan Goel also argued on X that models, inference, harness, context and deployment tools should be co-designed. The platform includes telephony, knowledge bases whose context is prefilled dynamically so the agent does not pause to search, tool calls into CRMs and backend APIs, and personalized outbound call batches.
Benchmarks
All figures are vendor-reported in Cartesia's blog, with GPT-5.5 as the reasoning model: a 1115 Elo and #1 of 30 agents on Speech Agent Arena, and 87.4% accuracy and 90.7% experience on EVA-Bench, both the highest in the table. EVA-Bench is ServiceNow's open benchmark, and its product lead Nitin Aggarwal gave a positive quote in the blog. Rankings will move as competitors update, and results with other reasoning models were not published. Many promotional review posts on X around the launch add little evidence.
Getting started
On the Agents page at play.cartesia.ai you can start from a template such as customer support, write instructions, pick a voice and model, and attach a phone number. Launch pricing is $0.03 per minute (50% off; no end date was given). Test with your own real call scripts, including interruptions, handoff to a human and failed tool calls, before comparing latency and cost with your current cascade.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.