NVIDIA GTC 2026 Open Models Panel Highlights with Jensen Huang

2026-03-25 · Source: NVIDIA · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Robotics & Autonomous Systems · Depth: Advanced, long

Summary

The discussion highlights a paradigm shift in AI, moving beyond the "proprietary versus open" debate to embrace a "proprietary and open" ecosystem. It introduces the concept of a "third type" of AI company that integrates both proprietary API-based models and custom open models to create specialized vertical products. A key development is the rise of "compound agents" and the "Perplexity computer" concept, which orchestrates multiple models and tools, treating AI as a system rather than just a model. This system aims to delegate complex tasks, leveraging different models' strengths without vendor lock-in. The content also addresses misconceptions, asserting that open models are not inherently behind frontier models and that fundamental knowledge infrastructure "yearns for openness," akin to historical shifts in publishing and science. The importance of post-training in model development is emphasized, along with the idea that while proprietary models may be generalists, open models excel as specialists, driving significant value.

Key takeaway

For CTOs and AI Product Managers evaluating AI strategy, recognize that a hybrid approach combining proprietary and open models, orchestrated by advanced agent systems, offers superior control, customization, and cost efficiency. Prioritize building robust orchestration layers to integrate diverse models, ensuring resilience and avoiding vendor lock-in, especially for mission-critical applications where trust and introspection are paramount.

Key insights

AI's future lies in a "proprietary and open" ecosystem, leveraging compound agents and orchestrated multi-model systems.

Principles

• AI is a system, not just a model.
• Openness is inevitable for fundamental knowledge infrastructure.
• Specialized models drive most value.

Method

Build an orchestration system (Perplexity computer) that integrates multimodal, multimodel, and multi-cloud capabilities, delegating tasks to diverse models based on their strengths.

In practice

• Combine proprietary and open models for vertical solutions.
• Focus on post-training for model development.
• Develop compound agents for complex, multi-day tasks.

Topics

• Proprietary and Open AI
• AI Agents
• AI Systems Orchestration
• Open Models
• AI Infrastructure

Best for: CTO, VP of Engineering/Data, AI Product Manager, AI Architect, Director of AI/ML, AI Engineer

Related on AIssential

• not much happened today — AI performance increasingly depends on context pipelines and open harnesses, not just model weights.
• OpenAI, Anthropic fight on the frontier — Leading AI labs are rapidly advancing models with self-improvement, multi-agent capabilities, and deep enterprise integration.
• Architecting Modern AI Systems: Platforms, Agents, and Integration — Open-source models and sovereign AI platforms offer cost control, data residency, and output flexibility over closed-source APIs.
• not much happened today — AI development is shifting towards multi-cloud distribution, cost-aware agentic workflows, and specialized hardware for inference.
• The Future of AI Is Open and Proprietary — AI's future is a hybrid ecosystem of open and proprietary models, orchestrated into specialized systems.
• Open and closed models are on different exponentials — Open and closed AI models follow divergent economic exponentials, creating distinct market structures and value capture mechanisms.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA.