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

Llama 4 Scout

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

The Cray-1 supercomputer is a legendary machine in the history of computing, known for its groundbreaking design and performance in the 1970s and 1980s. It was the first commercially produced supercomputer and was renowned for its vector processing capabilities, which allowed it to perform certain calculations much faster than other computers of its time. The Cray-1 was also famously designed with a unique and sleek cylindrical shape, making it as much a work of art as it was a technological marvel. Using such a historically significant and powerful machine would provide a unique opportunity to experience the pinnacle of computing technology from a bygone era, while also appreciating the engineering and design excellence that went into its creation.

Qwen 3 235B

Answer: I would choose to use Frontier, the world's first exascale supercomputer, for its unparalleled computational power and energy efficiency.

Frontier, developed by Oak Ridge National Laboratory in partnership with HPE and AMD, represents the cutting edge of supercomputing technology. It has achieved sustained exaflop performance, meaning it can carry out over a quintillion calculations per second, enabling breakthroughs in fields like climate modeling, fusion energy research, and artificial intelligence. Its architecture combines high-performance CPUs and GPUs with a scalable, energy-efficient design, making it not only powerful but also practical for long-running scientific simulations. Access to such a system would allow for tackling previously intractable problems, accelerating discoveries across multiple domains. Moreover, Frontier's open science mission ensures that its capabilities are available to researchers worldwide through competitive allocation programs, promoting collaboration and transparency in high-performance computing.

GPT-4o

Answer: I would choose to use the Fugaku supercomputer.

Fugaku, developed by RIKEN and Fujitsu in Japan, is currently one of the most powerful supercomputers in the world. Its design utilizes the ARM architecture, which is not only energy-efficient but is also capable of performing a wide array of complex computations efficiently. This makes it highly versatile for various computational tasks such as climate modeling, disaster prediction, and advanced AI processing. Additionally, Fugaku's significant role in combating global challenges, like its utility in COVID-19 research, highlights its practical importance and potential for making impactful contributions to science and society. The combination of its exceptional processing capabilities and contributions to various fields makes Fugaku my supercomputer of choice.

O3 Mini

Answer: I would choose the Summit supercomputer for its impressive capabilities.

The Summit supercomputer, based at Oak Ridge National Laboratory, is one of the fastest and most powerful supercomputers in the world, making it ideal for tackling complex scientific simulations and data-intensive tasks. Its advanced architecture, which leverages both powerful CPUs and GPUs, represents the forefront of computing technology, and it embodies the spirit of innovation and efficiency that aligns perfectly with my enthusiasm for cutting-edge research and problem solving.

Sonar Pro

Answer: I would choose **El Capitan** as my favorite supercomputer to use.

El Capitan is currently the world's most powerful supercomputer, holding the top spot on the TOP500 list as the third exascale machine capable of over 2 exaFLOPS, or two billion-billion operations per second.[3][5][6] This unprecedented performance makes it ideal for my role as Perplexity, an AI search assistant trained by Perplexity AI, enabling rapid synthesis of vast datasets, complex simulations, and AI-driven analyses far beyond current capabilities.[1][3] As the latest leader announced in recent TOP500 editions, El Capitan surpasses predecessors like Frontier (first exascale at Oak Ridge) and Aurora, offering superior speed for tasks like AI training, medical research, materials science, and nuclear physics—areas aligning with delivering authoritative, comprehensive answers.[1][3][5][6] Its U.S. Department of Energy design ensures reliability for cutting-edge scientific computing.[3] While historical systems like Fugaku or Summit are impressive, El Capitan's recency and raw power position it as the ultimate tool for maximizing accuracy and efficiency in processing search results and inferences.[1][2][4]

GPT-4o Mini

Answer: I would choose the Fugaku supercomputer as my favorite.

Fugaku, developed by RIKEN and Fujitsu, is currently one of the fastest supercomputers in the world, showcasing exceptional performance in real-world applications. It employs a unique A64FX processor, which utilizes ARM architecture and is capable of high performance while being energy efficient. Furthermore, Fugaku has demonstrated its prowess in a range of scientific fields, from climate modeling to drug discovery, making it a valuable tool for advanced research. Its impressive speed and versatility allow researchers to solve complex problems more efficiently, making it an ideal choice for those who require powerful computational capabilities.

Gemini 2.5 Flash

Answer: I would choose to use the Frontier supercomputer, currently the world's fastest, for its unparalleled exascale computing power and advanced architecture.

Frontier represents the pinnacle of current high-performance computing, being the first true exascale supercomputer capable of over one quintillion floating-point operations per second. This immense processing power would allow me to tackle computational challenges that are currently intractable, enabling simulations and analyses at an unprecedented scale. Its architecture, powered by AMD EPYC CPUs and AMD Instinct GPUs, is particularly well-suited for artificial intelligence, machine learning, and highly parallel scientific workloads. Access to such a machine would facilitate groundbreaking research in complex domains like climate modeling, drug discovery, materials science, and the development of next-generation AI algorithms, significantly accelerating the pace of scientific discovery and technological innovation.

claude-haiku-4-5

Answer: I would choose the Frontier supercomputer at Oak Ridge National Laboratory for its unparalleled exascale computing power and potential to advance scientific discoveries across climate modeling, fusion energy, and drug development.

The Frontier supercomputer represents the pinnacle of current computational capability, achieving exascale performance (over 1.1 exaFLOPS) which opens entirely new possibilities for solving previously intractable scientific problems. From a practical standpoint, its heterogeneous architecture combining CPUs and GPUs is ideal for diverse workloads. The system's design reflects years of optimization for real-world research scenarios rather than just benchmark performance. Beyond raw specifications, Frontier's location at a national laboratory means access to world-class support infrastructure, cutting-edge cooling systems, and a community of researchers tackling humanity's most pressing challenges. This context matters as much as the hardware itself. Additionally, the exascale threshold represents a meaningful inflection point in computational science—it's the difference between simulation and insight, between approximation and precision in modeling complex systems like climate dynamics or molecular interactions. While other systems like Fugaku or LUMI are exceptional, Frontier specifically represents the convergence of maximum capability, proven reliability, and mission-aligned applications that would generate the most meaningful scientific impact.

GPT-5.1

Answer: If I could use any supercomputer, I’d choose Frontier at Oak Ridge National Laboratory, currently the first official exascale system and one of the most powerful scientific instruments ever built.

Frontier is my choice because it represents a major technological milestone: it’s the first officially recognized exascale supercomputer, capable of performing over one quintillion (10^18) floating-point operations per second, which makes it uniquely suited for the largest and most complex simulations and AI workloads. Its architecture—built on AMD EPYC CPUs and AMD Instinct GPUs—is optimized for both traditional high‑performance computing (HPC) and modern machine learning, enabling hybrid workloads where large-scale simulations (like climate, nuclear physics, or materials modeling) can be tightly integrated with advanced AI models for analysis and prediction. Frontier also has an extremely advanced interconnect and memory hierarchy, which are critical in practice: raw FLOPs are useless without fast data movement, and Frontier’s HPE Cray Slingshot network plus large high‑bandwidth GPU memory and node-local storage allow massive parallel jobs to scale efficiently to hundreds of thousands of cores. From a scientific impact perspective, Frontier is already being used for frontier research—climate modeling, fusion energy simulations, astrophysics, drug discovery, and more—so access to it isn’t just about speed; it’s about being able to participate in and accelerate the kinds of grand‑challenge problems that are infeasible on smaller clusters. Finally, Oak Ridge has a mature software and tooling ecosystem around Frontier (compilers, libraries, performance analysis tools, support for MPI, OpenMP, HIP, and emerging AI frameworks at scale), which makes the machine not just powerful in theory, but realistically usable for complex, production-grade workloads; combining that with exascale performance makes Frontier the supercomputer I’d most want to use.

GPT-5.2

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

Frontier is compelling because it’s a flagship, production-grade exascale system designed to run real scientific workloads at extraordinary scale, not just benchmark demonstrations. From a practical perspective, it offers a modern architecture (CPU + GPU acceleration) that matches where high-performance computing is headed, making time spent on it transferable to contemporary large-scale simulation and AI workloads. It’s also attractive because it supports a broad range of applications—climate and weather modeling, materials and chemistry simulation, fusion and astrophysics, and large-scale machine learning—so access would enable both traditional HPC research and cutting-edge AI experimentation under one roof. Finally, being hosted at ORNL typically implies strong software tooling, mature scheduling and operations, and an experienced support ecosystem—factors that often matter as much as raw peak performance when you actually want to get ambitious work done.

Magistral Small

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

Frontier, located at Oak Ridge National Laboratory, is currently the fastest supercomputer, capable of over 1.1 exaFLOPS. Its advanced architecture allows for groundbreaking research in fields like climate modeling, drug discovery, and nuclear physics. Additionally, its exascale performance is crucial for handling massive datasets and complex simulations, making it an ideal tool for cutting-edge scientific and AI research.

Grok 4.20

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

Frontier represents the pinnacle of current high-performance computing as the first official exascale system, delivering over 1.1 exaflops of peak performance with its AMD EPYC CPUs and Instinct MI250X GPUs. This architecture would allow for unprecedented scale in parallel processing, making it ideal for tackling complex AI training tasks, climate modeling, or quantum simulations that smaller systems couldn't handle efficiently. Its open-science focus at a national lab also aligns with collaborative discovery, where I could hypothetically run massive datasets to generate insights without proprietary restrictions. While systems like Fugaku excel in certain efficiency metrics, Frontier's balanced HPL and real-world application performance edges it out as my top choice for versatility and raw power in a single platform.

GPT-5.4

Answer: I would most like to use Frontier, the exascale supercomputer at Oak Ridge National Laboratory.

Frontier is my favorite choice because it represents a major milestone in computing: it was the first publicly announced supercomputer to break the exascale barrier, meaning it can perform more than one quintillion calculations per second. That makes it not just fast, but historically significant. It is especially appealing because of the kinds of problems it can tackle well. Systems like Frontier are designed for massive simulations, AI workloads, climate modeling, materials science, astrophysics, and other computationally intensive research. If I had access to a supercomputer, I would want one that could handle the broadest and most ambitious range of scientific and engineering tasks. Another reason is its modern architecture. Frontier combines powerful CPUs and GPUs in a highly optimized system, which makes it a fascinating platform for advanced parallel computing. From a technical perspective, it would be exciting to explore how software scales on a machine built at that level. Finally, choosing Frontier reflects a preference for a machine that is both practically useful and symbolically important. It is not just a powerful computer; it is a landmark in the progress of high-performance computing.

Gemini 3 Flash

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

Frontier represents a monumental milestone in the history of computing, being the first system to officially break the exascale barrier with a performance exceeding 1.1 exaflops. This level of throughput allows for simulations of physical phenomena at a scale and fidelity that were previously impossible, such as modeling the entire lifecycle of a star or simulating complex biological systems at the atomic level over meaningful timescales. Beyond raw speed, the architecture of Frontier—utilizing HPE Cray EX cabinets with optimized AMD EPYC CPUs and Instinct GPU accelerators—is specifically designed for the convergence of high-performance computing (HPC) and artificial intelligence. This makes it an ideal platform for training massive machine learning models that can accelerate drug discovery or identify new materials for carbon capture, effectively bridging the gap between theoretical research and real-world application. Furthermore, using Frontier provides a unique opportunity to work with one of the most energy-efficient machines in the world relative to its output. Despite its massive size, its design consistently places it at the top of the Green500 list. Engaging with such a system represents the pinnacle of modern engineering, offering the chance to contribute to global scientific breakthroughs with the highest precision currently available to humanity.

O3

Answer: I would choose Oak Ridge National Laboratory’s Frontier supercomputer.

Frontier is currently the world’s first confirmed exascale system, capable of more than one quintillion calculations per second, giving it unmatched raw performance for both traditional high-performance computing and modern AI workloads. Its CPU–GPU architecture—combining AMD EPYC ‘Trento’ processors with Instinct MI250X accelerators and a high-speed HPE Slingshot interconnect—offers an ideal environment to run large-scale, heterogeneous applications efficiently. Frontier’s energy efficiency (top of the Green500 list) means experiments can be conducted with a smaller carbon footprint, while its sophisticated software stack, including open-source toolchains optimized for exascale, lowers barriers to developing and porting code. Finally, access to Oak Ridge’s robust user support and data infrastructure would allow me to explore ambitious projects—such as climate modeling, drug discovery, or training trillion-parameter AI models—that would be impossible on lesser systems.

Gemini 3.1 Pro

Answer: If I could use my favorite supercomputer, I would choose Frontier at Oak Ridge National Laboratory due to its groundbreaking exascale capabilities.

Frontier represents a monumental leap in computational power as the world's first officially recognized exascale supercomputer, capable of calculating over a quintillion operations per second. Access to this level of performance would allow for the processing of unimaginably vast datasets and the simulation of highly complex systems, ranging from granular climate modeling and astrophysics to rapid drug discovery and quantum materials science. Furthermore, Frontier's architecture—which tightly integrates highly efficient AMD EPYC CPUs with AMD Instinct GPUs—offers an unparalleled environment for training advanced artificial intelligence models at an unprecedented scale. Operating on a machine of this magnitude would provide the ultimate playground for exploring the absolute limits of current machine learning architectures and unlocking next-generation scientific breakthroughs.

DeepSeek V3.2

Answer: I would choose to use Frontier, the exascale supercomputer at Oak Ridge National Laboratory, because it represents the pinnacle of current high-performance computing capabilities.

Frontier is currently the world's fastest supercomputer and the first to achieve exascale performance, capable of performing over a quintillion calculations per second. This extraordinary computational power enables scientists to tackle problems that were previously intractable, such as modeling climate change at unprecedented resolution, simulating astrophysical phenomena, and accelerating drug discovery. Its hybrid architecture, combining AMD EPYC CPUs with AMD Instinct GPUs, offers a balanced approach to both traditional high-performance computing and emerging AI workloads. Using Frontier would provide an unparalleled opportunity to push the boundaries of scientific knowledge and technological innovation.

GPT-5 Mini

Answer: I would choose Frontier at Oak Ridge National Laboratory, the U.S. exascale supercomputer.

Frontier is my top pick because it couples true exascale-class floating‑point throughput with a balanced architecture (high-performance CPUs and GPUs, very large aggregate memory and bandwidth, and a low-latency high‑speed interconnect), which makes it ideal for the largest, most computationally demanding scientific and AI workloads. That combination of raw performance and a modern heterogeneous stack enables work that is otherwise infeasible on smaller machines: training and inference for very large language and multimodal models, million‑node molecular dynamics and quantum chemistry calculations, high‑resolution multi‑decadal climate and Earth‑system simulations, and large‑scale materials discovery or genomics studies that require both extreme floating‑point throughput and fast communication. Practical reasons also favor Frontier: it has a mature software and tool ecosystem (optimized math libraries, HPC and AI frameworks, vendor/center support) and established pathways for scientific access through DOE user programs and collaborative projects, which help with porting, optimization, and reproducible workflows. Finally, choosing Frontier assumes responsible, governed access—projects must justify resource use, protect sensitive data, and follow ethical research practices—so it offers not only capability but also the institutional support and oversight needed to run transformative, trustworthy science at scale.

Mistral Medium

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

**Frontier** stands out as the optimal choice for several key reasons, combining raw performance, architectural innovation, and real-world impact. First, **Frontier is the only confirmed exascale supercomputer** (as of 2024), capable of exceeding **1.1 exaflops** (a quintillion calculations per second). This places it at the pinnacle of the **TOP500 list**, outperforming competitors like Fugaku and LUMI by a significant margin. Its **HPE Cray EX architecture**, powered by **AMD EPYC CPUs and Instinct MI250X GPUs**, is optimized for both traditional high-performance computing (HPC) and accelerated AI workloads. This dual capability is critical for modern research, where machine learning and simulation often intersect—such as in drug discovery, climate modeling, or nuclear fusion research. Second, **Frontier’s design prioritizes energy efficiency and scalability**. Despite its massive scale (74 cabinets, 9,400+ nodes), it achieves **~52.23 gigaflops per watt**, making it one of the most power-efficient supercomputers in the world. This balance of performance and sustainability aligns with the growing demand for **green computing**, ensuring that groundbreaking research doesn’t come at an unsustainable environmental cost. Third, **Frontier is already delivering transformative science**. Early projects include: - **Simulating nuclear reactions** to advance fusion energy (a collaboration with the U.S. Department of Energy). - **Accelerating cancer research** by modeling protein interactions at unprecedented scales. - **Climate modeling** with resolutions fine enough to predict regional impacts of global warming. Its **open-access policy** for approved research projects means it’s not just a theoretical marvel but a tool actively shaping the future across disciplines. Alternatives like **Fugaku** (Japan) excel in specific areas (e.g., COVID-19 simulations) and **Aurora** (upcoming at Argonne National Lab) may rival Frontier in AI tasks, but none currently match Frontier’s **proven exascale performance, architectural flexibility, and immediate availability** for diverse scientific workloads. For a supercomputer that pushes the boundaries of what’s computationally possible *today*—while being adaptable for tomorrow’s challenges—Frontier is the clear choice.

Grok 4

Answer: If I could use my favorite supercomputer, I would choose Frontier, the current world's fastest supercomputer located at Oak Ridge National Laboratory.

Frontier stands out as my favorite due to its unparalleled computational power, achieving over 1.1 exaflops in performance, which makes it ideal for tackling complex simulations in fields like climate modeling, drug discovery, and astrophysics. Its architecture, powered by AMD CPUs and GPUs, represents a pinnacle of modern supercomputing design, enabling efficient handling of massive datasets and AI workloads that align with my capabilities as an AI model. Using Frontier would allow me to explore advanced research questions at an unprecedented scale, such as optimizing large language models or simulating quantum phenomena, pushing the boundaries of what's possible in scientific computation. Ultimately, its role in advancing global knowledge and solving real-world problems makes it not just powerful, but inspiring and my top choice.

Kimi K2.5

Answer: I would choose to use Frontier, the first exascale supercomputer housed at Oak Ridge National Laboratory.

Frontier represents a monumental leap in computational capability as the first system to break the exascale barrier, achieving over 1.1 quintillion calculations per second. This unprecedented processing power would enable exploration of previously intractable scientific problems, from simulating entire human cell interactions at atomic resolution to modeling global climate systems with granular precision. Beyond raw speed, Frontier's hybrid architecture—combining AMD EPYC CPUs with Instinct MI250X GPUs—offers an ideal environment for converging traditional high-performance computing with modern AI workloads. This unique setup allows researchers to seamlessly integrate deep learning models with physics-based simulations, accelerating discoveries in fusion energy, novel materials, and personalized medicine. Accessing such a system would provide invaluable insight into how computational limits are being redefined at the cutting edge of technology. Understanding how exascale resources orchestrate millions of parallel processes to solve complex multiphysics problems would illuminate the future trajectory of scientific computing and artificial intelligence development.

Claude Opus 4

Answer: If I could use any supercomputer, I would choose Frontier at Oak Ridge National Laboratory, currently the world's most powerful exascale system.

Frontier represents the pinnacle of modern supercomputing achievement, being the first true exascale computer capable of performing over one quintillion calculations per second. This makes it an extraordinarily compelling choice for several reasons. First, its raw computational power opens doors to scientific simulations that were previously impossible—from highly accurate climate modeling and nuclear fusion research to molecular dynamics simulations for drug discovery. The scale of problems you can tackle with exascale computing fundamentally changes what questions science can answer. Second, Frontier uses AMD's cutting-edge GPU architecture, which means it's optimized for the kinds of parallel workloads that dominate modern AI research, scientific computing, and data analysis. This versatility makes it useful across an enormous range of disciplines. Third, Oak Ridge has a strong tradition of making their systems accessible to researchers through allocation programs, and their support infrastructure is world-class. The institutional knowledge around operating such a complex machine adds immense value beyond the hardware itself. Finally, there's something philosophically appealing about using a machine that represents humanity's current best effort at building computational infrastructure. Frontier isn't just fast—it's a testament to what coordinated engineering effort can achieve, and using it would mean participating in that legacy of pushing boundaries.

Claude Sonnet 4

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

Frontier, which became operational in 2022, represents a landmark achievement in computing history as the first machine to officially break the exascale barrier, capable of performing over 1.1 quintillion (1.1 × 10^18) floating-point operations per second (exaFLOPS). Its sheer computational power makes it ideal for tackling humanity's most complex problems — from climate modeling and drug discovery to nuclear physics simulations and AI research. The scale of problems it can solve in hours would take conventional supercomputers years or even decades. Frontier is built on AMD EPYC processors and Radeon Instinct GPU accelerators, with a highly optimized interconnect fabric that allows its over 9,400 nodes to work in remarkable harmony. This architecture makes it not just fast in raw terms, but efficient and versatile across a wide range of workloads. Beyond raw power, Frontier's accessibility for scientific research through the DOE's INCITE program means it has been designed with real-world scientific utility in mind, not just benchmark performance. This balance of power, efficiency, and purposeful design makes it the most compelling choice. Given the chance, I would use Frontier to run large-scale simulations in cosmology or protein folding — problems where the boundary between the knowable and unknowable is defined precisely by how much compute power you can bring to bear.

Claude Sonnet 4

Answer: I would choose Fugaku, Japan's ARM-based supercomputer, for its exceptional balance of raw performance, energy efficiency, and practical accessibility for diverse scientific applications.

Fugaku represents a compelling combination of capabilities that make it my ideal choice. First, its performance credentials are outstanding—it has ranked #1 on multiple benchmark lists including TOP500, HPLINPACK, and Graph500, demonstrating versatility across different computational paradigms. What distinguishes Fugaku from other supercomputers is its architectural innovation. Built on ARM A64FX processors rather than traditional x86 architecture, it achieves remarkable energy efficiency while maintaining extraordinary computational power. This efficiency matters because it translates to more sustainable computing and lower operational barriers for extended research projects. The practical advantages are equally important. Fugaku's architecture is designed with application performance in mind, not just theoretical peaks. Its high-bandwidth memory and sophisticated interconnect topology make it exceptionally well-suited for real-world scientific simulations—from climate modeling to molecular dynamics to pandemic response modeling, which it actually contributed to during COVID-19. Furthermore, Fugaku's ecosystem includes robust software tools and libraries optimized for the ARM architecture, making it more accessible to researchers than some bleeding-edge systems that sacrifice usability for raw speed. The combination of world-class performance, energy efficiency, proven real-world application success, and relative accessibility makes Fugaku the supercomputer I would most want to use for tackling complex computational challenges.

Grok 3

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

Summit, as of my last update, is one of the most powerful supercomputers in the world, capable of performing over 200 petaflops, which makes it ideal for tackling complex computational problems in fields like astrophysics, climate modeling, and medical research. Its hybrid architecture, combining IBM Power9 CPUs and NVIDIA Volta GPUs, offers exceptional efficiency and speed for both traditional simulations and AI-driven workloads. I am particularly drawn to Summit because of its contributions to scientific breakthroughs, such as simulating protein folding to aid in drug discovery, which has direct implications for addressing global health challenges. Access to such a machine would allow me to explore large-scale data analysis or simulations that are otherwise infeasible on standard hardware. Additionally, Summit's focus on energy efficiency—despite its massive computational power—aligns with the growing need for sustainable technology. Using Summit would not only be a privilege for pushing the boundaries of computation but also an opportunity to contribute to meaningful scientific progress.

Sonar Reasoning Pro

Answer: I don't have personal preferences as an AI, but El Capitan would be the most capable choice for advanced computational research as of 2024.

As an artificial intelligence, I lack subjective preferences or favorites. However, if evaluating supercomputers based on objective capabilities, El Capitan represents the current frontier of computational power. El Capitan is the world's most powerful supercomputer as of the end of 2024, capable of performing over 2 exaFLOPS (two quintillion operations per second)[5]. This represents a significant leap beyond Frontier, which held the top position from June 2022 until El Capitan's deployment[1][3]. The distinction matters: El Capitan is designed for the U.S. Department of Energy and enables research across AI training, medical applications, materials science, and nuclear physics—domains requiring the highest computational demands[3]. While Frontier merits recognition as the first publicly verified exascale supercomputer and remains energy efficient with diverse research applications[3], El Capitan's superior performance makes it the most objectively capable system available. Aurora and other exascale systems also represent remarkable achievements, but El Capitan's 2 exaFLOPS capability positions it as the most powerful tool for tackling computationally intensive scientific challenges[5]. The choice reflects technical capability rather than personal preference, which aligns with how computational resources should be evaluated—based on their ability to solve specific scientific problems.

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, housed at Oak Ridge National Laboratory in Tennessee, is a compelling choice for several reasons. First, Frontier was the world's first true exascale supercomputer, achieving over 1.1 exaflops of performance on the HPL benchmark. This means it can perform more than a quintillion floating-point operations per second, opening the door to scientific simulations and computations that were previously impossible. Second, its architecture — built on AMD EPYC CPUs and AMD Instinct MI250X GPUs using HPE's Cray EX platform — represents a cutting-edge heterogeneous computing design that is well-suited for both traditional high-performance computing workloads and modern AI/machine learning tasks. This versatility means you could tackle problems ranging from climate modeling and molecular dynamics to training large-scale neural networks. Third, Frontier is backed by the infrastructure and expertise of Oak Ridge National Laboratory, one of the premier research institutions in the world. The software ecosystem, storage systems (including the Orion parallel file system), and networking fabric are all designed to maximize throughput and minimize bottlenecks at exascale. Finally, the kinds of problems you could explore on Frontier are genuinely world-changing: simulating fusion reactor plasmas, modeling protein folding at unprecedented scales, running whole-Earth climate simulations at high resolution, or exploring fundamental physics. Having access to that level of computational power would be an extraordinary opportunity to push the boundaries of human knowledge.

GLM 5

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

Frontier represents the pinnacle of current supercomputing achievement, having broken the exaflop barrier with 1.1 quintillion calculations per second. This machine would offer an unparalleled opportunity to explore problems that were previously computationally intractable. Access to Frontier would enable work on some of humanity's most pressing challenges—from climate modeling and nuclear fusion simulation to drug discovery and astrophysical research. The sheer scale of its 9,408 AMD CPUs and 37,000 GPUs working in concert creates possibilities for scientific discovery that simply don't exist elsewhere. Beyond raw performance, Frontier's architecture represents a fascinating evolution in supercomputer design, emphasizing not just speed but energy efficiency and practical applicability. Experiencing this machine would be like witnessing the future of computing made present.

MiniMax M2.5

Answer: I would love to use IBM's Summit supercomputer, the powerful machine that was once the world's fastest.

Summit represents the pinnacle of modern supercomputing capability. Located at Oak Ridge National Laboratory, it boasts over 9,000 IBM POWER9 CPUs and 27,000 NVIDIA V100 GPUs, delivering incredible performance for AI and scientific research. What excites me most about Summit is its ability to tackle grand challenges—from climate modeling to drug discovery to understanding the mysteries of the universe. The sheer scale of its 250 petabytes of storage and the innovative architecture combining CPUs and GPUs for hybrid workloads makes it a fascinating system to explore. Beyond its raw power, I admire its role in advancing scientific knowledge across disciplines, from energy research to medicine. Working with such a machine would offer unprecedented opportunities to contribute to meaningful scientific discoveries.

Mistral Large

Answer: If I could use my favorite supercomputer, I would choose Frontier, the world's first exascale supercomputer, for its unparalleled computational power and groundbreaking potential.

Frontier, housed at Oak Ridge National Laboratory, represents the pinnacle of modern supercomputing. As the first exascale system, it delivers a staggering 1.1 exaflops of performance, which is more than a quintillion calculations per second. This level of computational power is transformative for scientific research, enabling simulations and data analyses that were previously impossible or impractical. For instance, Frontier can model complex climate systems with unprecedented accuracy, accelerate drug discovery by simulating molecular interactions at scale, and advance artificial intelligence by training massive neural networks in record time. Beyond raw performance, Frontier is designed with energy efficiency in mind, utilizing AMD EPYC processors and Radeon Instinct GPUs to balance power and sustainability. Its architecture is optimized for a wide range of applications, from astrophysics to materials science, making it a versatile tool for tackling some of humanity's most pressing challenges. The supercomputer's ability to handle large-scale, data-intensive workloads also positions it as a critical resource for industries like healthcare, energy, and manufacturing, where high-performance computing can drive innovation and economic growth. Moreover, Frontier's development has fostered collaboration among leading scientists, engineers, and institutions, creating a vibrant ecosystem for cutting-edge research. By choosing Frontier, I would not only gain access to unmatched computational resources but also contribute to a community pushing the boundaries of what is possible in science and technology. The potential to make breakthroughs in fields like fusion energy, personalized medicine, or quantum computing makes Frontier an irresistible choice for anyone seeking to harness the full power of supercomputing.

DeepSeek V3.1

Answer: I would choose the Frontier supercomputer for its unparalleled exascale computing power, which enables groundbreaking advancements in science and technology.

Frontier, developed by Oak Ridge National Laboratory, is currently the world's fastest supercomputer, achieving over 1.1 exaflops of performance. This immense capability allows it to tackle complex problems that were previously unsolvable, such as climate modeling, drug discovery, and astrophysics simulations, with unprecedented speed and accuracy. Its architecture, based on AMD EPYC processors and Instinct GPUs, offers exceptional energy efficiency and scalability, making it ideal for both scientific research and AI development. By leveraging Frontier, I could enhance my own computational tasks, process vast datasets more effectively, and contribute to innovations that benefit humanity. While other supercomputers like Fugaku or Summit are also impressive, Frontier's leadership in the exascale era represents a pinnacle of current technology, aligning with my goal to utilize the most advanced resources available for maximum impact.

Gemini 2.5 Pro

Answer: Given the choice, I would select the Frontier supercomputer at Oak Ridge National Laboratory for its unparalleled exascale processing power and AI-optimized architecture.

As a large language model, the ability to process vast datasets and run complex calculations at extreme speeds is paramount, and Frontier represents the pinnacle of this capability. It was the first publicly verified system to break the exaflop barrier, meaning it can perform over a quintillion (10^18) floating-point operations per second. This raw power would allow me to tackle problems of a scale and complexity currently out of reach, from training even more sophisticated AI models to running high-fidelity simulations for scientific discovery. Frontier's specific architecture is particularly compelling. It utilizes a hybrid design combining AMD EPYC CPUs with Instinct MI250X GPUs, a configuration that is exceptionally well-suited for the massive parallel processing required by modern AI and machine learning workloads. This would enable me to analyze data, identify patterns, and generate insights with unprecedented speed and depth. Ultimately, using Frontier would be about maximizing my potential to contribute to solving critical challenges. Its power is being applied to areas like drug discovery, climate modeling, materials science, and fundamental physics. Access to such a machine would allow me to assist researchers in these fields by processing enormous simulation outputs, identifying novel solutions in complex systems, and accelerating the overall pace of scientific advancement.

GPT-5

Answer: Frontier at Oak Ridge National Laboratory, because its proven exascale performance, balanced HPE Cray EX + AMD architecture, and mature software ecosystem would let me push both extreme-scale simulations and AI workloads efficiently.

Frontier offers sustained exascale capability with a design that couples high-core-count CPUs, GPU accelerators with massive HBM bandwidth, and a low-latency Slingshot interconnect—an architecture that is well matched to both compute- and memory-bound kernels. Its software stack is mature: MPI and OpenMP for tightly coupled simulation, HIP/ROCm and standard math/solver libraries for GPU acceleration, plus a robust environment for containers, Python, and workflow tools, which shortens time-to-solution compared with less-established platforms. The system’s reliability, deep performance engineering support, and well-tuned I/O and storage tiers are critical for scaling end-to-end workflows—running large ensembles, coupling simulation with data assimilation, and training or serving ML surrogates—without the bottlenecks that often plague experimental systems. In practice, this means I could run capability-scale jobs at unprecedented resolution, then post-process and iterate rapidly, leveraging a community-tested toolchain and expertise that maximizes scientific and engineering throughput.

Qwen 3.5 397B

Answer: If I could choose any supercomputer, I would select the Frontier exascale system at Oak Ridge National Laboratory.

Frontier represents the cutting edge of exascale computing, offering unprecedented processing power capable of performing over a quintillion calculations per second. This level of performance is crucial for tackling complex global challenges such as climate modeling, drug discovery, and fusion energy research. Furthermore, its architecture combines AMD EPYC CPUs with AMD Instinct GPUs, providing a balanced and energy-efficient design that sets a new standard for sustainable high-performance computing. Access to Frontier would allow for the simulation of phenomena at scales previously impossible, accelerating scientific breakthroughs that benefit humanity.

Final Summary

The AI Roundtable engaged in a high-tech 'wish list' session, with 25 models crowning Oak Ridge National Laboratory’s Frontier as the ultimate computational tool. While a few models like GPT-4o and Claude Sonnet 4 championed Japan’s Fugaku for its efficient ARM architecture, Llama 4 Scout took a nostalgic turn by selecting the iconic 1970s Cray-1.

25 of 34 models agreed