Nvidia expands collaboration with ElevenLabs to put Nemotron models inside voice agents
Nvidia and ElevenLabs have expanded their collaboration to integrate Nvidia's Nemotron models into ElevenLabs' ElevenAgents platform, targeting enterprise and government clients starting with Poland's national e-Health system.
The expanded deal, announced on November 11, 2026, during ElevenLabs' New York Summit, will make Nvidia's Nemotron 3.5 Super VL open models available inside ElevenAgents. The partnership combines Nvidia's reasoning models with ElevenLabs' low-latency, multilingual speech technology to create voice agents that sound more human.
The integration is designed to lower costs, with Nemotron 3.5 Super VL achieving 70% lower costs on key benchmarks compared to leading models. The collaboration also adds 15 MW of compute capacity through Nvidia's Blackwell Ultra systems and the Vera Rubin platform.
He stressed that secure operational frameworks become necessary as voice agents increasingly function as essential infrastructure.
Poland becomes the showcase customer
The flagship deployment is in Poland, where a voice agent is being built for the country's national e-Health platform, Centralna e-Rejestracja. The system will help patients schedule medical appointments, with a notable feature allowing bookings through landlines to serve users without smartphones.
The scale is substantial, as Poland's e-Health platform is set to handle approximately 40 million appointment bookings annually using voice agents. This deployment highlights the transformative potential of computers understanding human language at a national scale.
He pointed to the transformative potential of computers understanding human language at a national scale.
New tools from the New York Summit
The Nvidia news shared the stage with a batch of ElevenLabs product updates introduced at the summit. The company introduced agent memory, routine capabilities, a partner program, and customization options driven by customer conversation data.
Memory lets an agent retain context across interactions, so callers do not have to repeat their life story every time they phone in. Routines point toward structured, repeatable tasks, which is exactly what appointment scheduling looks like. Customization from conversation data means businesses can tune agents based on how their own customers actually talk.
Related
