GPT-5.6 Sol Automates Quantum Experiments at MIT

OpenAI’s flagship AI model, GPT-5.6 Sol, is being used by researchers at MIT to automate critical aspects of quantum computing experiments, significantly streamlining workflows in one of the most complex fields of modern science. Beatriz Yankelevich, a graduate student with MIT’s Engineering Quantum Systems Group (EQuS), has successfully deployed the model to autonomously conduct measurements on superconducting qubits, the building blocks of quantum processors.

Quantum computing relies on qubits, which operate using the principles of quantum mechanics. Unlike classical bits, qubits can exist in superpositions of states, allowing quantum computers to tackle computational problems that are intractable for traditional systems. However, preparing and calibrating qubits is a labor-intensive process requiring hundreds to thousands of interdependent measurements—a problem MIT researchers are addressing with AI.

Yankelevich connected GPT-5.6 Sol to laboratory software via Codex, enabling the AI to execute routine measurement workflows autonomously. The system analyzes results, adjusts experimental parameters dynamically, and saves findings for subsequent stages. This automation has freed researchers from constant supervision, allowing them to focus on higher-value tasks like designing experiments and interpreting data. “I can have agents running measurements overnight or while I’m working in the cleanroom,” Yankelevich said. “I can check in from my phone, see what they’ve done, and steer them if needed.”

The EQuS team tested GPT-5.6 Sol on a six-qubit chip, a standard configuration used to benchmark fabrication processes. The AI autonomously identified qubit transition frequencies, calibrated control pulses, and determined quantum coherence times—tasks that would typically take researchers several days. While the system excelled with clear experimental signals, it struggled with noisy or weak data, occasionally requiring human intervention. This indicates that while AI is proficient in defined workflows, interpreting ambiguous physical results remains a challenge.

OpenAI’s GPT-5.6 Sol was first previewed on June 26, 2026, and became generally available alongside its counterparts, Terra and Luna, on July 9, 2026. Sol stands out as the most capable model in the GPT-5.6 family, designed for tasks requiring deep reasoning and adaptability. Beyond research, it has applications in coding, cybersecurity, and other high-complexity domains. To encourage adoption, OpenAI recently reduced API pricing for Sol by over 20% on August 21, 2026, making it more accessible to researchers and developers.

For MIT, integrating GPT-5.6 Sol into quantum research represents a paradigm shift in how experiments are conducted. The AI doesn’t just automate; it collaborates, enabling multiple agents to tackle different problems simultaneously. This not only accelerates research but also expands the scope of what’s possible in experimental design. “I’ve built infrastructure to guide agents through measurement, theory, and chip design, and now it’s really starting to pay off,” Yankelevich explained.

While AI like GPT-5.6 Sol won’t replace human expertise, its ability to augment capabilities, save time, and reduce repetitive workloads underscores its transformative potential. As quantum computing moves closer to practical applications, tools like Sol could become indispensable in bridging the gap between theoretical research and real-world implementation.

Image source: Shutterstock