Closing the Gap: Can Open Models Match Frontier Performance and How Would You Even Know?

The distance between open and closed models continues to shrink on paper, but matching the frontier means something different depending on who is asking. For some, it is a research question about how far the right data and fine-tuning can push a model; for others, it is a practical question about whether a model they can own and run is finally good enough to trust. Either way, the challenge is not just making the claim, it is proving it. Static benchmarks increasingly favor models that are fluent, confident, and wrong, and the gaps that matter most rarely show up on a leaderboard. This panel will explore what it actually takes to close that gap and the rigorous, head-to-head evaluation needed to determine whether you have truly closed it, or only think you have.