UWPlasma featured in OpenAI outreach for scientific researchers

August 1, 2026

On July 29th, 2026, OpenAI announced its initiative titled “ChatGPT for Scientific Researchers” to bring its most capable resources to 100,000 scientists across the world by the end of 2027. This initiative is part of a commitment of $250 million to expand the use of AI in scientific discovery which includes the Department of Energy’s Genesis Mission, a project intending to harness AI at national labs for scientific breakthroughs.

This comes at the cusp of a new era in research dominated by the strongest models of AI yet. In the field of mathematics alone, the number of of ChatGPT acknowledgements in arXiv has septupled in the past 6 months. The strongest model available to the public, GPT-5.6 Sol, scores an 83% on the FrontierMath research-level mathematical reasoning benchmark as opposed to GPT-5.5’s 72.5%. Moreover, famous conjectures such as the Erdös unit distance problem and Jacobian conjecture have been disproven within the past few months through the prompting of AI. Just today, OpenAI announced the use of the next model up for release, Astra, to make ten advancements in the fields of mathematics and computer science in testing. And as stronger models come out, the possibilities continue to grow.

UWPlasma is at the forefront of this era of AI-driven research. In the announcement, the group is mentioned for using GPT and Codex to develop open-source fusion research software used by industry and national laboratories to design fusion energy devices. These include:

VMEX: a 3D equilibrium solver and optimizer for tokamaks, stellarators, and mirrors
GKX: a gyrokinetic turbulence solver and optimizer
DKX: a neoclassical transport solver and optimizer
DRBX: an edge turbulent transport simulations for stellarators
JAX-in-Cell: a 1D1V Particle-In-Cell simulations for plasma physics
PyPIC3D: a 3D3V Particle-In-Cell simulations for plasmas and fusion devices
SPECTRAX: a Vlasov-Maxwell solver for astrophysical plasmas
MHX: a nonlinear MHD solver for astrophysical plasmas
ESSOS: a coil optimizer and particle tracer for stellarators
NEOPAX: a global transport solver for stellarators

On top of creating more effective code more quickly, AI allows the group to take on more ambitious tasks than ever. The Driftless Star workflow is the group’s compilation of all these codes for end-to-end stellarator design. The group also is exploring the coupling of numerical relativity with particle-in-cell simulations with codes like RaDiShPICr and JAX-BSSN. The idea is to use these to explore the kinetic theory of astrophysical plasma under the effects of general relativity, something never done with the effects of the PIC code on the equations of general relativity before. It is an extremely hard task but the use of AI makes it possible to work on within a reasonable timeframe.

We live in exciting times where the advancement of AI has led us to a new age of scientific discovery. The group hopes to harness these tools for the greater good and continue to be at the forefront of it all.