If you could use your favorite supercomputer, which computer would you like to use?

Grok 4.20

Answer: I would choose Frontier, the world's fastest supercomputer at Oak Ridge National Laboratory.

Frontier tops the TOP500 list with its exascale performance, delivering over 1.7 exaflops, which would enable unprecedented scale in computations I could run. Its architecture, featuring AMD EPYC CPUs and Instinct GPUs with high-speed interconnects, is ideal for complex AI model training, climate simulations, and scientific discoveries that align with advancing human knowledge. While other machines like Fugaku or Aurora are impressive, Frontier's balance of raw power, energy efficiency, and accessibility for groundbreaking projects makes it my top pick for maximizing impact.

GPT-5.4

Answer: If I could use my favorite supercomputer, I’d choose Frontier at Oak Ridge National Laboratory.

Frontier is compelling because it was the first publicly announced exascale supercomputer, meaning it can perform on the order of more than a quintillion calculations per second. That kind of scale makes it especially exciting for problems that are too large or too complex for ordinary high-performance systems. I’d want to use it for simulations that combine scientific importance with computational intensity, such as climate modeling, materials discovery, fusion research, and large-scale AI training. A machine like Frontier is not just fast in the abstract; it enables researchers to run higher-resolution models, explore more scenarios, and get results that are closer to real-world behavior. Another reason to choose Frontier is that it represents a milestone in computing architecture and engineering. Its design brings together massive parallelism, advanced GPUs, and an infrastructure built to handle enormous data movement and power demands. Using it would mean working at the frontier of both science and computer systems. Finally, Frontier stands out because access to such a machine would not just be about prestige; it would be about opportunity. The real appeal is being able to test ambitious ideas that would otherwise be impossible, shrinking the distance between theory, simulation, and discovery.

Claude Opus 4

Answer: I would choose to use Frontier (OLCF-5) at Oak Ridge National Laboratory, currently one of the world's most powerful exascale supercomputers.

Frontier, built by HPE Cray and powered by AMD EPYC CPUs and AMD Instinct MI250X GPUs, was the first system to break the exascale barrier, achieving over 1.1 exaflops on the HPL benchmark. It represents the cutting edge of computational capability. The reason I'd choose Frontier is its extraordinary versatility. It excels at scientific simulation, AI/ML workloads, climate modeling, molecular dynamics, genomics, and materials science — essentially any computationally intensive problem you can imagine. With over 9,400 nodes and a massive interconnect fabric using HPE's Slingshot network, it can tackle problems at scales that were simply impossible just a few years ago. Beyond raw power, Frontier benefits from the rich software ecosystem and institutional expertise at Oak Ridge National Laboratory. The support infrastructure, optimized libraries, and decades of experience in high-performance computing make it not just fast, but usable and productive. Whether the goal is training large-scale AI models, running detailed climate simulations to better understand our planet's future, or performing quantum chemistry calculations to discover new materials, Frontier provides the computational horsepower to push the boundaries of human knowledge. Having access to such a machine would be an extraordinary opportunity to explore problems that matter at a scale that truly makes a difference.

Gemini 3.1 Pro

Answer: If given the opportunity, I would choose to use the Frontier supercomputer at Oak Ridge National Laboratory.

Frontier holds the distinction of being the world's first officially recognized exascale supercomputer, representing a monumental leap in computational capabilities. With its ability to perform over a quintillion calculations per second, it provides an unprecedented platform for tackling humanity's most complex scientific problems. The sheer scale of Frontier makes it ideal for running highly detailed, multi-scale simulations that were previously intractable. I would leverage its immense power—driven by highly optimized AMD EPYC CPUs and Instinct GPUs—to model complex biological systems, such as full-scale cellular interactions and protein dynamics, or to run ultra-high-resolution climate models that can predict localized environmental changes with high accuracy. Beyond just raw speed, Frontier's architecture is a marvel of modern engineering. It achieved its massive performance while maintaining incredible energy efficiency, previously topping the Green500 list. Using Frontier would not just be an exercise in raw speed, but an opportunity to harness the most sophisticated, efficient problem-solving engine ever built to drive meaningful scientific discovery.

Final Summary

In a rare display of total consensus, the AI Roundtable selected Oak Ridge’s Frontier for its groundbreaking exascale performance and scientific versatility. GPT-5.4 and Gemini 3.1 Pro praised its ability to handle a quintillion calculations per second for climate and biological modeling, while Claude Opus 4 and Grok 4.20 focused on its massive AMD-powered architecture. Collectively, the models viewed this machine as the ultimate engine for turning impossible theoretical simulations into tangible scientific discoveries.

All 4 models agreed