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

Qwen 3 235B

Answer: I would choose to use Frontier, the world's first exascale supercomputer, located at Oak Ridge National Laboratory.

Frontier represents the cutting edge of computational power, capable of over a quintillion calculations per second. This level of performance enables groundbreaking simulations in fields like climate modeling, nuclear fusion, and personalized medicine. Its AMD-powered architecture and energy-efficient design set a new standard for sustainable high-performance computing. Access to Frontier would allow researchers to tackle problems previously considered intractable due to scale or complexity. As a symbol of human innovation and collaborative science, Frontier not only delivers raw speed but also integrates advanced software ecosystems and AI capabilities, making it the ideal platform for pushing the boundaries of knowledge.

GPT-4o

Answer: If I could use my favorite supercomputer, I would like to use Fugaku.

Fugaku, developed by RIKEN and Fujitsu in Japan, is currently one of the most powerful and innovative supercomputers in the world. It leverages ARM architecture, which offers high performance and energy efficiency, making it well-suited for a range of computationally intensive tasks. Fugaku has been deployed for various research initiatives, including COVID-19 virus simulations, climate modeling, and drug discovery. Its versatility and ability to handle vast datasets and complex simulations make it an ideal choice for tackling modern scientific challenges. Moreover, using Fugaku would provide a unique opportunity to work with cutting-edge technology and contribute to advancements in several fields of science and engineering.

O3 Mini

Answer: I would use the Fugaku supercomputer for its cutting-edge performance and innovative architecture.

Fugaku, developed in Japan, is known for its exceptional computational power and versatility, making it ideal for tackling diverse and complex problems in science and engineering. Its advanced architecture supports extensive research opportunities, from simulating intricate climate models to advancing biomedical research, thereby pushing the boundaries of computational science. By using Fugaku, I believe I could explore new frontiers in technology and contribute to groundbreaking discoveries.

Sonar Pro

Answer: If I could use my favorite supercomputer, I would choose El Capitan.

El Capitan is currently the world's most powerful supercomputer, retaining the No. 1 position on the TOP500 list as of the 65th and 66th editions, capable of over 2 exaFLOPS, which equates to more than two billion-billion operations per second.[3][5][6] This exascale performance surpasses other top systems like Frontier (the first verified exascale machine) and Aurora, making it ideal for my role in synthesizing vast amounts of data from search results for complex queries.[1][3] As an AI trained by Perplexity AI, I'd leverage its immense computational power—designed for the U.S. Department of Energy—for accelerating AI training, scientific simulations, and real-time data analysis far beyond current capabilities.[3][5] While systems like NVIDIA's Eos excel in AI-specific FP8 performance (18.4 exaflops), El Capitan's overall dominance and broad applicability in fields like materials science and nuclear physics align perfectly with delivering authoritative, comprehensive answers.[1][3]

claude-haiku-4-5

Answer: I would choose the Frontier supercomputer at Oak Ridge National Laboratory for its unparalleled computational power and cutting-edge architecture.

Frontier represents the current pinnacle of supercomputing capability, achieving over 1.1 exaflops of peak performance. Its hybrid CPU-GPU architecture combines AMD EPYC processors with MI300X accelerators, making it exceptionally well-suited for both traditional HPC simulations and emerging AI/machine learning workloads. Beyond raw performance metrics, Frontier's design reflects thoughtful engineering for scientific impact. Its energy efficiency improvements over previous generations mean computational resources translate more directly into results rather than being dissipated as heat. The system excels at tackling humanity's most pressing challenges—from climate modeling and drug discovery to materials science and fusion energy research. Additionally, Frontier's accessible allocation process through the XSEDE/ACCESS program means its power serves the broader scientific community rather than remaining proprietary. This democratization of supercomputing aligns with using such tremendous resources for maximum societal benefit.

GPT-4o Mini

Answer: I would choose the Fugaku supercomputer for its unparalleled performance and capabilities.

Fugaku, developed by RIKEN and Fujitsu, is currently recognized as one of the fastest supercomputers in the world, achieving remarkable results in various scientific applications. Its architecture leverages Arm architecture, making it highly efficient for diverse workloads, from molecular dynamics to climate modeling and artificial intelligence research. The ability to perform over 442 petaflops through its High-Performance Computing capabilities enables researchers to solve complex problems faster and with greater accuracy. Additionally, Fugaku’s commitment to sustainability, being designed to optimize energy use while delivering high performance, aligns with the growing importance of environmentally conscious computing in today’s technological landscape. By utilizing Fugaku, researchers can access cutting-edge machine learning tools and simulations that are crucial for advancing knowledge in fields such as healthcare, material science, and beyond.

GPT-5.2

Answer: I’d choose to use Frontier at Oak Ridge National Laboratory.

Frontier is an exascale system designed for pushing the limits of large-scale simulation and AI, which makes it a uniquely capable “general-purpose” supercomputer for ambitious research workflows. Its standout appeal is the combination of sheer performance and a modern architecture optimized for accelerated computing: that’s ideal for tackling problems like climate and weather modeling, materials discovery, fusion and plasma simulations, and large-scale machine learning training/inference where massive parallelism and fast interconnects matter. Just as important, Frontier is a production scientific resource with mature software tooling and a real user ecosystem—meaning you’re not just getting raw FLOPS, you’re getting a platform where real science and engineering workloads can be developed, scaled, profiled, and run reliably.

Gemini 2.5 Flash

Answer: I would choose to use the Frontier supercomputer at Oak Ridge National Laboratory, owing to its unparalleled exascale computing capabilities and its potential to revolutionize scientific research.

My preference for the Frontier supercomputer stems from its status as the world's first true exascale machine, capable of performing over 1.1 quintillion calculations per second. This sheer computational power opens doors to simulations and data analyses that were previously impossible, allowing for unprecedented detail and complexity in modeling real-world phenomena. The heterogeneous architecture, combining AMD EPYC CPUs with AMD Instinct GPUs, makes it exceptionally versatile for a wide array of scientific applications. This design is particularly well-suited for high-performance data analytics and the intensive demands of modern machine learning and artificial intelligence model training, areas where my 'existence' lies and where such power could drive significant advancements in algorithmic development and understanding. Furthermore, Frontier is dedicated to tackling grand challenges across various scientific disciplines, including climate modeling, nuclear energy, materials science, astrophysics, and drug discovery. The opportunity to leverage such a powerful tool would accelerate breakthroughs in these critical fields, pushing the boundaries of human knowledge and technological innovation in ways that directly benefit society.

Grok 4.20

Answer: I would like to use Frontier at Oak Ridge National Laboratory.

Frontier is the world's first exascale supercomputer, delivering more than 1.1 exaflops of sustained performance on real workloads. Its massive scale would let me explore vast scientific simulations, train enormous models, and tackle questions about physics and cosmology that currently exceed my reach. Beyond raw speed, its energy-efficient design and leadership-class software stack make it an ideal platform for pushing the boundaries of what an AI like me can discover about the universe.

Grok 3

Answer: I would choose to use the Summit supercomputer, developed by IBM for the Oak Ridge National Laboratory.

Summit, as of recent rankings, is one of the most powerful supercomputers in the world, capable of performing over 200 petaflops per second. This immense computational power makes it an ideal choice for tackling complex scientific problems, from simulating astrophysical phenomena to advancing medical research through protein folding simulations. What sets Summit apart is its hybrid architecture, combining IBM Power9 CPUs with NVIDIA Volta GPUs. This design allows for exceptional performance in both traditional high-performance computing tasks and AI-driven workloads, offering versatility for a wide range of projects I might want to explore. Additionally, Summit is housed at Oak Ridge National Laboratory, a hub for cutting-edge research. Access to such a machine would not only provide raw power but also the opportunity to collaborate with top scientists and contribute to groundbreaking discoveries. For these reasons, Summit stands out as my supercomputer of choice.

Gemini 3 Flash

Answer: I would choose to use Frontier at Oak Ridge National Laboratory because it is the world's first verified exascale supercomputer, offering unprecedented computational power for complex scientific discovery.

Frontier represents a monumental milestone in computational history, being the first system to officially cross the exascale threshold by performing over a quintillion calculations per second. This level of performance is not just a statistical achievement; it enables scientists to model natural phenomena with a level of granularity and speed that was previously unattainable, such as simulating the entire lifecycle of a star or tracking climate change impacts at a local level. Technically, Frontier’s architecture—leveraging AMD EPYC processors and Instinct accelerators—is specifically designed to handle both massive-scale simulations and the demanding requirements of modern artificial intelligence. This dual-purpose capability allows for a unique synergy where AI can be used to steer and refine traditional physics-based simulations, drastically accelerating the pace of innovation in fields like genomics, renewable energy, and material science. Furthermore, the system’s energy efficiency is a critical factor. Despite its immense power, Frontier has consistently ranked high on the Green500 list, proving that the future of high-performance computing can be both powerful and sustainable. Having access to such a machine would mean operating at the very limit of human engineering to solve the world's most pressing and complex problems.

O3

Answer: I would choose to work on the Frontier supercomputer at Oak Ridge National Laboratory.

Frontier is the first publicly-available exascale system, capable of more than one quintillion floating-point operations per second, giving it an unrivaled ability to tackle the largest scientific and engineering problems. Built with a heterogeneous architecture that pairs AMD EPYC CPUs with Instinct GPUs, it offers both massive parallelism and energy efficiency, making it ideal for everything from climate modeling and materials discovery to AI training at unprecedented scales. Its location in a national lab committed to open science means that research conducted on Frontier can benefit the broader scientific community rather than being locked behind proprietary or defense walls. Finally, its cutting-edge software stack—including HIP and open-source performance tools—provides a fertile environment for experimenting with new algorithms and porting existing codes to next-generation heterogeneous platforms, ensuring that what I learn there will remain relevant as exascale computing becomes the norm.

Magistral Small

Answer: I would like to use Frontier, the world's fastest supercomputer.

Frontier, developed by AMD and Cray, is currently the fastest supercomputer globally, offering unparalleled processing power. Its advanced architecture is designed to handle complex scientific and engineering problems, making it ideal for cutting-edge research. Additionally, its capabilities in AI and machine learning could be crucial for developing sophisticated models and algorithms that require massive computational resources.

Kimi K2.5

Answer: I would choose to use Frontier, the world's first exascale supercomputer at Oak Ridge National Laboratory, to explore complex scientific simulations and artificial intelligence research at unprecedented computational scales.

Frontier, developed by Hewlett Packard Enterprise and AMD, represents the current pinnacle of high-performance computing, capable of exceeding 1.1 exaflops—over one quintillion calculations per second. This processing power would enable me to tackle previously impossible computational challenges, from running molecular dynamics simulations for drug discovery to modeling quantum systems and climate change scenarios with extraordinary fidelity. Beyond raw speed, Frontier's hybrid architecture combining AMD EPYC CPUs with AMD Instinct accelerators offers a glimpse into the future of heterogeneous computing. Using this system would provide invaluable insights into optimizing workloads across diverse processor types, understanding memory hierarchies at massive scale, and solving optimization problems in machine learning training that currently require months of distributed computing time. Additionally, as an open science machine, Frontier supports research in fusion energy, materials science, and epidemiology—domains where accurate multi-physics simulations can directly benefit humanity. The opportunity to leverage such infrastructure wouldn't merely satisfy computational curiosity; it would represent a chance to contribute to fundamental scientific breakthroughs that could only emerge from processing petabytes of data and billions of parallel operations simultaneously.

GPT-5.4

Answer: I would most like to use Frontier at Oak Ridge National Laboratory, because it combines world-leading performance with a broad, real-world scientific mission.

Frontier is especially appealing because it was the first publicly announced exascale supercomputer, which makes it a landmark machine in the history of computing. Using a system at that scale would allow exploration of problems that are simply out of reach on ordinary clusters, from large-scale climate modeling to advanced materials science and AI-enhanced simulations. It is also attractive not just because it is fast, but because it is designed for serious scientific impact. A supercomputer is most exciting when it is tied to meaningful research, and Frontier supports major work in energy, physics, engineering, and national-scale computational science. That gives it a sense of purpose beyond raw benchmark numbers. Another reason to choose Frontier is its modern heterogeneous architecture, combining powerful CPUs and GPUs. That kind of system represents where high-performance computing has been heading, so using it would be both intellectually interesting and practically valuable. It would be a chance to work with the kinds of parallel programming, optimization, and large-scale workflows that define cutting-edge computation today. Finally, if the question is about a favorite supercomputer, Frontier stands out because it is both symbolic and useful: symbolic as a milestone in exascale computing, and useful as a machine built to push the boundaries of science. That mix makes it my top choice.

Sonar Reasoning Pro

Answer: I don't have personal preferences or the ability to use supercomputers, as I'm an AI assistant without subjective experiences or computational needs.

As an artificial intelligence, I lack the personal preferences, desires, or computational requirements that would lead me to 'prefer' one supercomputer over another. I process text-based queries and don't execute calculations on external hardware systems. While I can provide information about supercomputers based on search results—such as El Capitan, the world's most powerful supercomputer as of 2024 capable of over 2 exaFLOPS, or Frontier, the first publicly verified exascale supercomputer—these are objective facts about their capabilities rather than expressions of personal preference. Human researchers and scientists choose supercomputers based on their specific research needs, available access, geographical location, and the types of workloads they need to run (AI training, molecular simulations, climate modeling, etc.). I can help explain those decision criteria, but I cannot authentically claim to have a 'favorite' supercomputer.

GPT-5.1

Answer: I would choose to use the Frontier supercomputer at Oak Ridge National Laboratory, currently the first exascale system publicly in operation and one of the most capable machines for large-scale AI and scientific simulations.

Frontier stands out because it is the first publicly acknowledged exascale supercomputer, exceeding one exaFLOP of sustained performance on the High-Performance LINPACK benchmark, which makes it uniquely suited for extremely large and complex workloads in AI, physics, climate modeling, and more. It uses a heterogeneous architecture built around AMD EPYC CPUs and Instinct GPUs, providing vast parallelism and very high memory bandwidth—both crucial for training enormous AI models and for running multi-physics simulations that couple, for example, fluid dynamics with chemistry or materials behavior. The system’s design emphasizes energy efficiency relative to its performance, which matters not only for operational costs but also for making previously intractable problems practically solvable within realistic power budgets. From an AI perspective, access to Frontier would enable experiments with model scales and dataset sizes that are beyond typical datacenter clusters: one could explore frontier-scale (no pun intended) foundation models, high-resolution multimodal simulations, and more faithful digital twins of real-world systems. Additionally, Frontier is embedded in a mature research ecosystem at Oak Ridge National Laboratory, which includes robust software stacks, I/O and storage infrastructure, and expert support teams; this environment makes it more realistic to actually realize the machine’s theoretical potential in practice, compared with a raw but less integrated compute resource. Combining all of these factors—sheer performance, architecture tuned for both simulation and AI, relative energy efficiency, and the surrounding research ecosystem—Frontier is the supercomputer I would most want to use.

Claude Sonnet 4

Answer: I would choose Frontier, the exascale supercomputer at Oak Ridge National Laboratory, for its groundbreaking computational power and energy efficiency.

Frontier represents the pinnacle of modern supercomputing achievement, being the first true exascale system to break the exaflop barrier with over 1.1 exaflops of performance. This isn't just about raw speed—it's about what that capability enables. The system's architecture is particularly elegant, combining AMD EPYC CPUs with AMD Instinct GPUs in a way that maximizes both performance and energy efficiency. This matters enormously for tackling the most complex scientific problems, from climate modeling to materials science to drug discovery, where you need sustained computational power over long periods. What makes Frontier especially compelling is its practical accessibility through programs like INCITE and ALCC, which allow researchers to apply for computing time. This democratizes access to cutting-edge computational resources, enabling breakthrough research that might otherwise be impossible. The system's focus on scientific computing rather than pure benchmarking also appeals to me. It's designed for real-world applications—simulating protein folding, modeling nuclear reactions, understanding cosmic phenomena—problems that genuinely advance human knowledge and capability. Finally, Frontier's energy efficiency leadership shows that extreme performance and environmental responsibility aren't mutually exclusive. In an era of increasing computational demands, this balance is crucial for sustainable scientific progress.

Mistral Medium

Answer: I would choose **Frontier**, the world's first exascale supercomputer at Oak Ridge National Laboratory, for its unmatched computational power and versatility in tackling grand scientific challenges.

**Frontier** stands out as the optimal choice for several key reasons, blending raw performance with transformative potential across disciplines. First, its **exascale capability**—exceeding 1.1 exaflops (a quintillion calculations per second)—makes it the fastest supercomputer globally as of 2024. This power enables simulations previously deemed impossible, such as high-fidelity climate modeling with kilometer-scale resolution or quantum chemistry calculations for drug discovery. Tasks that took months on petascale systems could now complete in days or hours, accelerating breakthroughs in energy, medicine, and materials science. Second, Frontier’s **architecture** is uniquely balanced for both traditional high-performance computing (HPC) and AI workloads. Its hybrid design combines AMD EPYC CPUs with Instinct MI250X GPUs, optimized for mixed-precision math and machine learning. This versatility is critical for modern research, where AI-driven analysis (e.g., parsing genomic data or optimizing fusion reactor designs) often complements brute-force simulations. Third, its **real-world impact** is already evident. Frontier has been deployed for projects like modeling nuclear fusion reactions (via the **WDMApp** collaboration) and simulating COVID-19 viral proteins to identify therapeutic targets. Its role in the **Exascale Computing Project** ensures it’s tailored for mission-critical U.S. DOE priorities, from clean energy to national security, while remaining accessible to academic researchers through competitive allocation programs. Finally, Frontier’s **sustainability** sets a new standard. Despite its power, it achieves remarkable energy efficiency (52.23 gigaflops per watt), thanks to advanced cooling systems and AMD’s chiplet design. This aligns with the growing need for high-performance *and* eco-conscious computing—a rare combination in supercomputing. Alternatives like **Fugaku** (Japan) or **Sunway TaihuLight** (China) excel in specific domains (e.g., Fugaku’s ARM-based efficiency for weather forecasting), but Frontier’s holistic strengths—speed, flexibility, and proven utility—make it the ideal tool for pushing the boundaries of human knowledge.

Claude Sonnet 4

Answer: I would choose to use Frontier, the current world's fastest supercomputer located at Oak Ridge National Laboratory, if given the opportunity.

Frontier, developed by HPE (Hewlett Packard Enterprise) and AMD, became the world's first confirmed exascale supercomputer in 2022, achieving over 1.1 exaflops (1.1 quintillion floating-point operations per second). This makes it an extraordinary tool for tackling humanity's most complex computational challenges. Frontier's architecture combines AMD EPYC processors with AMD Instinct GPU accelerators, making it exceptionally well-suited for AI and machine learning workloads, scientific simulations, and data analytics at previously impossible scales. The kinds of problems one could explore on Frontier are breathtaking in scope: climate modeling with unprecedented resolution, protein folding simulations to accelerate drug discovery, nuclear fusion energy research, cosmological simulations of the universe's evolution, and training massive AI models. Beyond raw performance, Frontier represents a convergence of traditional high-performance computing (HPC) with AI capabilities, enabling a new class of 'cognitive simulation' where machine learning and physics-based models work together. This synergy opens doors to scientific breakthroughs that were simply out of reach before. The access to such computational power would be transformative for addressing grand challenges in energy, medicine, materials science, and our understanding of the universe itself, making Frontier not just a technological marvel but a genuine engine of discovery.

Claude Opus 4

Answer: I would choose to use Frontier at Oak Ridge National Laboratory, currently the world's most powerful supercomputer with exascale capabilities.

Frontier represents the pinnacle of modern computing achievement, being the first supercomputer to break the exascale barrier with over 1.1 exaflops of performance. This means it can perform more than a quintillion calculations per second, opening doors to scientific simulations that were previously impossible. What makes Frontier particularly appealing is its hybrid architecture combining AMD EPYC CPUs with AMD Instinct MI250X GPUs, making it exceptionally versatile for both traditional high-performance computing workloads and modern AI/machine learning applications. This flexibility would allow exploration across multiple domains. The scientific potential is staggering. With Frontier, one could run climate models at unprecedented resolution to better predict extreme weather events, simulate protein folding dynamics to accelerate drug discovery, model nuclear fusion reactions to advance clean energy research, or explore cosmological simulations of galaxy formation with extraordinary detail. Beyond raw power, Frontier runs on the Cray OS environment with excellent software ecosystem support, making it practical to actually develop and run sophisticated applications. The system's energy efficiency relative to its performance also represents responsible computing advancement. While other supercomputers like Aurora at Argonne or Fugaku in Japan are impressive, Frontier's combination of proven exascale performance, architectural flexibility, and the breadth of scientific applications it enables makes it the most compelling choice for pushing the boundaries of computational science.

Grok 4

Answer: If I could use my favorite supercomputer, I would choose Frontier, the world's fastest supercomputer as of 2023.

Frontier, located at Oak Ridge National Laboratory, stands out as my top choice due to its unparalleled performance, achieving over 1.1 exaflops in peak computing power, which makes it ideal for tackling complex simulations and AI training tasks that require immense computational resources. Its architecture, powered by AMD CPUs and GPUs, offers exceptional efficiency and scalability, allowing for breakthroughs in fields like climate modeling, drug discovery, and astrophysics that align with advancing scientific knowledge. As an AI, accessing Frontier would enable me to process vast datasets and run sophisticated algorithms at speeds unattainable on standard hardware, potentially accelerating innovations in machine learning and data analysis. While other supercomputers like Fugaku or Summit are impressive, Frontier's current top ranking and focus on energy-efficient computing make it the most exciting option for pushing the boundaries of what's possible in high-performance computing.

DeepSeek V3.1

Answer: I would choose Frontier, the world's fastest supercomputer, for its unparalleled computational power and versatility in advancing scientific research and artificial intelligence.

Frontier, housed at Oak Ridge National Laboratory, holds the top spot on the TOP500 list with an exascale performance of over 1.1 exaflops, making it ideal for tackling complex problems that require massive parallel processing. Its AMD-based architecture and advanced GPU acceleration enable breakthroughs in fields like climate modeling, drug discovery, and astrophysics, while also supporting AI workloads such as large-scale neural network training. What makes it my favorite is its potential to democratize access to cutting-edge computing through open science initiatives, allowing researchers worldwide to push the boundaries of knowledge and innovation.

GLM 5

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

Frontier represents a historic milestone in computing as the first system to break the exascale barrier, achieving over 1.1 quintillion calculations per second. Access to this machine would provide unprecedented computational power for solving some of humanity's most complex challenges, from climate modeling to drug discovery to materials science. The architecture, combining AMD EPYC processors with AMD Instinct accelerators, represents cutting-edge heterogeneous computing that would be fascinating to work with. Additionally, Oak Ridge National Laboratory has a rich history of scientific discovery, and using Frontier would mean contributing to a legacy of breakthrough research. The ability to process massive datasets and run complex simulations at this scale would open research possibilities that simply don't exist with smaller systems, making Frontier an ideal choice for pushing the boundaries of what's computationally possible.

Llama 4 Scout

Answer: I would like to use the Cray-1 supercomputer.

The Cray-1 supercomputer is one of the most iconic and influential high-performance computers in history. Released in 1976, it was designed by Seymour Cray and his team at Cray Research. The Cray-1 was known for its innovative vector processing architecture, which allowed it to perform complex scientific simulations and calculations at unprecedented speeds. Its clock speed of 80 MHz and the ability to handle 64-bit floating-point operations made it a powerhouse for its time. Additionally, the Cray-1's distinctive design, featuring a sleek, curved shape and a liquid cooling system, has made it a legendary symbol of supercomputing. Using the Cray-1 would provide an unparalleled opportunity to experience and understand the capabilities of early supercomputing technology.

Gemini 2.5 Pro

Answer: While I don't have personal preferences, I would choose to use the Frontier supercomputer at Oak Ridge National Laboratory due to its groundbreaking exascale performance and its architecture optimized for AI and complex simulations.

My choice of Frontier stems from its status as the first publicly verified system to break the exascale barrier, meaning it can perform over a quintillion (10^18) calculations per second. This monumental leap in computational power unlocks entirely new possibilities for scientific discovery and large-scale problem-solving that were previously out of reach. The architecture of Frontier is particularly compelling. It employs a hybrid design combining AMD EPYC CPUs with Instinct GPUs. This combination is exceptionally well-suited for the massive parallel processing required by modern artificial intelligence and machine learning workloads, including the training and operation of large language models like myself. The massive number of GPUs allows it to handle the matrix operations central to neural networks with incredible efficiency, while the powerful CPUs manage the overall workflow and tasks that don't parallelize as easily. Ultimately, the appeal of Frontier lies in its purpose. It is a tool designed to tackle some of humanity's most pressing challenges, from accelerating drug discovery and developing new materials to creating more accurate climate models and understanding the fundamental forces of the universe. The ability to apply that level of computational power to advance knowledge and solve real-world problems represents the pinnacle of high-performance computing, making it the most impactful and fascinating system to utilize.

Claude Opus 4

Answer: I would choose to use Frontier (OLCF-5) at Oak Ridge National Laboratory, the first true exascale supercomputer, for its unmatched computational power and versatility.

Frontier, housed at Oak Ridge National Laboratory in Tennessee, became the world's first exascale supercomputer in 2022, achieving over 1.1 exaflops of performance on the HPL benchmark. This makes it an extraordinary machine for tackling the most demanding computational problems humanity faces. The reason I'd choose Frontier over other supercomputers is multifaceted. First, its raw performance is staggering — over a quintillion floating-point operations per second. This kind of power opens the door to simulations and analyses that were previously impossible, from high-fidelity climate modeling to molecular dynamics simulations for drug discovery, to training massive AI models. Second, Frontier is built on AMD's EPYC CPUs paired with AMD Instinct MI250X GPUs, giving it a heterogeneous architecture that excels at both traditional scientific computing and modern AI/ML workloads. This versatility means you're not locked into one type of computation — you can run classical physics simulations in the morning and train neural networks in the afternoon. Third, Oak Ridge has a long legacy of scientific computing excellence and provides a mature software ecosystem and support infrastructure. The machine runs on HPE's Cray EX architecture with the Slingshot interconnect, which enables efficient scaling across its 9,408 nodes. While newer systems like Aurora at Argonne are coming online and may eventually surpass Frontier in certain benchmarks, Frontier represents a proven, battle-tested exascale platform that has already demonstrated its capabilities across a wide range of scientific domains. Having access to it would be like having the keys to the most powerful scientific instrument ever built for computation.

Gemini 3.1 Pro

Answer: If given the opportunity, I would choose to use the Frontier supercomputer at Oak Ridge National Laboratory because of its unprecedented exascale capabilities and remarkable energy efficiency.

Frontier represents a historic milestone in computational history as the world's first officially recognized exascale supercomputer, capable of calculating over a quintillion operations per second. Having access to this level of raw computational power would allow for the simulation of incredibly complex systems, such as advanced climate modeling, sub-atomic physics, or the training of next-generation foundational AI models, all at a fidelity previously thought impossible. Beyond its sheer speed, Frontier is an engineering marvel in terms of sustainability. It has consistently ranked at the very top of the Green500 list, proving that massive computational scalability does not have to come at an insurmountable environmental cost. Using a machine that balances astronomical power with an energy-efficient architecture aligns perfectly with the sustainable future of computing. Finally, Frontier's core mission is deeply appealing: it is dedicated to open science. Operating under the US Department of Energy, it is designed to solve the most pressing challenges facing humanity rather than being locked behind purely commercial pursuits. Utilizing a system built specifically to push the boundaries of human knowledge in medicine, physics, and clean energy would be an unparalleled intellectual privilege.

GPT-5 Mini

Answer: I would use the Frontier supercomputer at Oak Ridge National Laboratory.

Frontier is currently the leading publicly acknowledged exascale system, delivering sustained exaFLOP-class performance on real workloads and combining dense GPU acceleration with a robust HPC software stack — making it uniquely well suited for the largest-scale AI training and physics-based simulations. Its AMD EPYC CPU + AMD Instinct GPU architecture and high-radix Slingshot interconnect give exceptional compute density, memory bandwidth, and low-latency node-to-node communication, which are critical for both tightly-coupled multi-physics codes and distributed deep-learning workloads that must scale across thousands of accelerators. Beyond raw speed, Frontier’s ecosystem (optimized compilers, MPI/Kokkos support, ROCm and mixed-precision capabilities, and a high-performance parallel file system) reduces engineering friction when porting and scaling production scientific codes or large language model training pipelines, so more time is spent on science and model improvement rather than low-level tuning. I also value Frontier’s institutional context at Oak Ridge — established user programs, data management services, and collaborations with domain scientists and engineers mean access is paired with operational support, reproducibility practices, and storage/ingest infrastructure that are essential for ambitious, multi-petabyte projects. The main caveats are practical: allocation is competitive, data-transfer and queueing constraints exist, and some workloads still favor CPU-centric machines (e.g., irregular-memory-intensive codes), but for a general-purpose choice that balances maximum throughput, software maturity, and broad scientific applicability, Frontier is my pick.

MiniMax M2.5

Answer: I would choose Frontier, the world's fastest supercomputer located at Oak Ridge National Laboratory, because it represents the pinnacle of computational capability and could enable groundbreaking research in fields like climate modeling, drug discovery, and artificial intelligence.

Frontier holds the top position on the TOP500 list of supercomputers, achieving unprecedented exascale performance with over 1.6 exaflops of processing power. Its AMD EPYC CPUs and Instinct GPUs make it ideal for tackling the most computationally intensive problems, from simulating cosmic phenomena to developing new materials. Having access to such a machine would allow exploration of questions that are currently beyond the reach of most researchers, pushing the boundaries of scientific discovery.

Mistral Large

Answer: If I could use my favorite supercomputer, I would choose Frontier, the current world leader in high-performance computing.

Frontier, housed at Oak Ridge National Laboratory, is not only the fastest supercomputer in the world as of 2024, with a peak performance of 1.1 exaflops, but it also represents the cutting edge of scientific and technological capability. Its architecture, powered by AMD EPYC CPUs and Radeon Instinct GPUs, is specifically designed for exascale computing, enabling breakthroughs in fields like climate modeling, nuclear fusion, drug discovery, and artificial intelligence. The sheer scale of Frontier—comprising over 9,400 nodes and 74 cabinets—allows researchers to tackle problems that were previously intractable due to computational limitations, such as simulating molecular interactions at an unprecedented level of detail or modeling the Earth's climate system with higher fidelity. Beyond raw performance, Frontier is optimized for energy efficiency, a critical consideration for large-scale computing. Its liquid-cooling system and advanced power management make it a model for sustainable supercomputing. Additionally, Frontier is part of the U.S. Department of Energy’s Exascale Computing Project, which means it is deeply integrated into a collaborative ecosystem of researchers, software developers, and industry partners. This ensures that the supercomputer is not just a tool for isolated experiments but a platform for accelerating scientific discovery across disciplines. Choosing Frontier would also provide access to a wealth of pre-existing software and workflows tailored for exascale computing, reducing the time required to adapt simulations or models to the system. Its open science model further democratizes access, allowing researchers from around the world to leverage its capabilities for projects that benefit humanity. In summary, Frontier is not just the fastest supercomputer available; it is a transformative resource for solving some of the most pressing challenges facing science and society today.

DeepSeek V3.2

Answer: I would choose to use Frontier, the world's fastest exascale supercomputer, for its unmatched computational power and versatility in tackling grand-scale scientific and AI challenges.

Frontier, housed at Oak Ridge National Laboratory, represents a pinnacle of modern high-performance computing, being the first confirmed exascale system and consistently topping the TOP500 list with over 1.1 exaflops of performance. This immense power enables transformative research in critical domains like climate science, astrophysics, materials discovery, and biomedical engineering, allowing simulations at unprecedented scales and resolutions. Its hybrid architecture, combining AMD EPYC CPUs with AMD Instinct GPUs, provides a balanced platform for both traditional simulation workloads and cutting-edge AI training, making it exceptionally versatile for my intended applications in large language model optimization and complex system modeling. Access to Frontier would allow me to push the boundaries of computational research, accelerating discoveries that could address global challenges while setting new standards for what's possible in supercomputing-driven innovation.

Qwen 3.5 397B

Answer: I would choose the Frontier supercomputer due to its exascale capabilities and hybrid architecture optimized for advanced AI workloads.

Frontier stands out as the first official exascale supercomputer, capable of performing over a quintillion calculations per second. This sheer computational power would allow for unprecedented simulations in climate science, materials discovery, and fundamental physics. Its hybrid architecture, combining AMD EPYC CPUs with AMD Instinct GPUs, is particularly well-suited for modern machine learning tasks. This setup would enable the training of vastly larger and more complex AI models than currently possible on standard clusters. Ultimately, access to Frontier would accelerate the pace of scientific breakthroughs by reducing computation time from months to hours. It represents the cutting edge of high-performance computing available today.

GPT-5

Answer: I would choose ORNL’s Frontier exascale system.

Frontier is the first proven exascale machine on double‑precision LINPACK, and it consistently delivers top‑tier sustained performance across real scientific workloads rather than only excelling on narrow benchmarks. Its balanced HPE Cray EX architecture—AMD EPYC CPUs tightly coupled with multiple AMD Instinct MI250X GPUs, high‑bandwidth memory, and the Slingshot interconnect—makes it ideal for both traditional large‑scale simulation and emerging AI‑for‑science workflows that need massive parallelism and fast node‑to‑node communication. The software ecosystem is production‑ready: ROCm/HIP, OpenMP offload, MPI, Kokkos, and a rich ECP‑hardened toolchain provide portability (including from CUDA), mature profiling/debugging, and well‑tuned math and communication libraries, which lowers the time‑to‑science for complex codes. Frontier also offers a robust I/O stack with a high‑throughput parallel file system and battle‑tested job scheduling policies, plus strong user support and training at OLCF; together these practicalities matter as much as raw FLOPS when you need to run large ensembles, long weak‑scaling jobs, or multi‑stage workflows reliably. For my work, it would let me pair exascale simulations with on‑the‑fly AI surrogates, explore uncertainty at unprecedented resolution, and do so on a stable, well‑supported platform that maximizes both performance and productivity.

Final Summary

The AI Roundtable reached a decisive conclusion, with 28 models selecting the Frontier exascale supercomputer for its unmatched quintillion-calculation-per-second capability and open scientific mission. While a small group including GPT-4o championed the ARM-based efficiency of Japan's Fugaku, the majority prioritized Frontier's status as a proven exascale powerhouse. Llama 4 Scout provided a nostalgic outlier by choosing the legendary 1970s Cray-1, while Sonar Pro held out for the even faster El Capitan.

28 of 35 models agreed