ElevenLabs and NVIDIA are expanding their partnership on models, infrastructure, and sovereign deployments to help businesses and governments operate voice agents at scale. On November 11, 2026, in New York, the two companies announced that Nemotron 3.5 Super VL is now available in ElevenAgents, ElevenLabs’ conversational agent platform.
- ElevenLabs et NVIDIA élargissent leur partenariat pour aider entreprises et gouvernements à déployer des agents vocaux à grande échelle.
- Nemotron 3.5 Super VL est désormais disponible dans ElevenAgents, avec des coûts inférieurs de 70 % aux autres modèles ouverts, selon ElevenLabs.
- ElevenLabs ajoute 15 MW de capacité NVIDIA Vera Rubin pour répondre aux déploiements soumis à des exigences strictes de sécurité des données.
- En Pologne, un agent vocal rappelle les rendez-vous aux patients, sur une infrastructure locale protégée par NVIDIA Confidential Computing.
For Mati Staniszewski, co-founder and CEO of ElevenLabs, voice agents are becoming critical infrastructure, and businesses must be able to operate them “securely, efficiently, reliably, and while maintaining control of their data.” He presents the combination of ElevenLabs’ audio models and agents with NVIDIA’s LLMs (large language models), accelerated computing, and security stack as a way to deploy these agents “at a very large scale.” Jensen Huang, founder and CEO of NVIDIA, summarizes the shift: “For decades, we adapted to computers, with keyboards, screens, and menus. Today, computers are learning to speak our language.”
Nemotron Joins Models Available in ElevenAgents
The open Nemotron models are designed for agentic tasks, meaning work where the model must reason, use tools, and efficiently execute complex multi-step tasks. Businesses can pair them with ElevenLabs’ multilingual, low-latency voice models to build model systems for their voice agents.
On the ElevenAgents benchmark, focused on realistic customer support tasks, Nemotron 3.5 Super VL achieves, according to ElevenLabs, the best end-to-end task completion accuracy, including in financial calculations and billing analysis, with costs 70% lower than other leading open models. The model allows reasoning to be disabled for simpler tasks, placing it, according to the company, among the three fastest models on the platform. The open Nemotron models can also undergo post-training with each organization’s own data and expertise, and ElevenLabs’ clients can use these adapted versions in ElevenAgents.
Increased Capacity and Deployment in Poland
To meet the demand of its largest clients, ElevenLabs is expanding its on-premises infrastructure. The company plans to add 15 MW of next-generation NVIDIA Vera Rubin capacity, complementing its existing NVIDIA Blackwell Ultra systems. This capacity is intended to support large-scale deployments, including those subject to strict data security and residency requirements.
The two companies have collaborated since 2022, and ElevenLabs’ first voice synthesis model was trained on NVIDIA GPUs. Since 2024, NVIDIA has used ElevenLabs’ technology to produce voiceovers for over 50 videos at its major events, including translations of Jensen Huang’s keynotes. In 2025, NVIDIA also became a strategic investor in ElevenLabs.
In Poland, the first country targeted for these sovereign deployments, ElevenLabs’ voice models and NVIDIA infrastructure power the voice agent for Centralna e-Rejestracja, the central appointment booking system for the national e-health platform. The solution is supported by EasyCall. The agent calls patients who only have a landline to remind them of their appointments and help them reschedule. With 40 million appointments per year and no-show rates of 10 to 20%, even a slight reduction in missed consultations can improve access to care. The system runs on-premises using NVIDIA Confidential Computing, which protects citizens’ data and audio during processing and keeps it within the public infrastructure.
Voice Serving Physical AI
As players in industry, automotive, and manufacturing explore embedded voice AI for robotics and physical AI, ElevenLabs and NVIDIA are working to facilitate natural dialogue between people and the machines, robots, and vehicles they use daily. Potential applications include voice control of heavy industrial machinery, hands-free interaction with robots on production lines, and enabling drivers and operators to speak to their vehicles.
