NVIDIA (NVDA) The Six Five Summit: AI Unleashed 2026 summary

Industry innovation and open models

• Open source acts as a catalyst for both open and closed innovation, driving rapid industry progress.

• The definition of "open" in AI models is nuanced, with varying degrees of openness in data, weights, and architecture.

• Open models have enabled a proliferation of use cases, especially when users can access and fine-tune models for specific needs.

• Value in AI models is increasingly seen as accruing at the data layer, not just in model weights or architecture.

• Enterprises benefit from customizing open models for in-domain tasks, achieving faster and more cost-effective solutions.

Multi-model strategies and enterprise adoption

• Organizations are adopting heterogeneous model strategies, combining open, closed, and frontier models based on use case.

• Edge AI and on-prem deployments are critical for use cases where latency and data sovereignty matter.

• Model routers and orchestrators like Switchyard enable dynamic selection of the best model for each workflow.

• Security and IP protection are addressed by running sensitive workloads on-prem with open models and using agent traces for governance.

• The ability to burst between local, cloud, and frontier models is now practical, with tools for PII redaction and policy-driven routing.

Accelerating innovation and the role of open source

• Fully open models, including datasets, lower barriers for enterprises and individuals to innovate and customize.

• Open source models and tooling accelerate post-training and evaluation, making AI more accessible.

• Real-world performance of open models can surpass benchmarks, especially when tailored with proprietary data.

• Harnesses, or the systems around models, are the new frontier for both usability and security in AI agents.

• Industry collaboration on security, such as the Shared AI Findings Exchange, is emerging to address risks in open and closed models.