Nvidia GTC 2026 to Host High-Stakes Debate on Open Frontier Models

Nvidia's annual GTC conference will convene the industry's most influential voices on open-source artificial intelligence for a panel discussion that could reshape the competitive landscape between proprietary and open frontier models.

Jensen Huang, Nvidia's CEO, will moderate a session titled "Open Models: Where We Are and Where We're Headed" featuring Mira Murati, founder and CEO of Thinking Machines Lab; Aravind Srinivas, CEO of Perplexity; Arthur Mensch of Mistral; Harrison Chase of LangChain; Misha Laskin of Reflection AI; and Robin Rombach of Black Forest Labs. The discussion is scheduled for March 18 from 12

The panel arrives at a pivotal moment for open-source AI development. According to the conference agenda, participants will evaluate "the open source market's progress, what's next for these emerging frontier models and potentially how they will shake up the AI market to compete against the leading proprietary models."

(The session represents one of five highlighted events at GTC 2026, which also includes presentations on Disney's robotics integration of animated characters through deep reinforcement learning and modular mechatronics.)

The timing reflects intensified scrutiny of open-source viability following DeepSeek's launch, which the conference materials cite as a catalyst for the current "thriving" state of open frontier models. The panel composition notably includes both pure-play open-source advocates and companies navigating hybrid commercial strategies, setting up potential friction over sustainability and competitive positioning against closed systems from OpenAI, Anthropic, and Google.

Murati's participation marks her first major public forum since departing OpenAI, where she served as chief technology officer and briefly as interim CEO. Her new venture, Thinking Machines Lab, has not yet disclosed its approach to model openness, making her perspective particularly anticipated. Srinivas leads Perplexity, which builds on open models while operating a proprietary search product, embodying the commercial tensions the panel will likely address.