NASA’s CEA v3 cuts 108,500-case runs to 1.11 seconds

A propulsion sweep can turn into a wall of waiting: NASA’s benchmark sent 108,500 cases through its modernized Chemical Equilibrium with Applications code, CEA v3, in approximately 1.11 seconds. The older CEA2 took approximately 15 minutes for the same workflow, according to NASA.

CEA is the software engineers use to calculate equilibrium chemistry, rocket performance, shock and detonation behavior. CEA2, released in 2002, remained widely used, but its procedural Fortran design lacked a subroutine interface—the connection that lets another program call the calculation directly. CEA v3 moves the code to Fortran 2008, with object-oriented structures, stricter typing and a thread-safe equilibrium solver.

That architecture is built for work that happens in batches: automated design studies, parametric sweeps, optimization and uncertainty analysis. Engineers can connect the code directly to Python, C, MATLAB or Excel, while a legacy command-line interface preserves existing CEA input files. NASA has also expanded the thermodynamic database with green-propellant constituents including ADN, HAN and LMP-103S, as well as n-Butanol, a sustainable aviation fuel candidate.

There is a trade-off. In representative testing, one CEA v3 equilibrium calculation was approximately 40 percent slower than a comparable CEA2 calculation. NASA puts the absolute difference at only about 0.004 seconds per case, while the multi-case benchmark showed an approximately 800-times reduction in runtime. Those figures come from NASA’s testing; no independent result is reported here.

So what changes in practice? Teams exploring many propulsion options can run those comparisons inside automated engineering workflows instead of stitching together separate tools or waiting through a long batch. The faster large-scale path makes broader design-space exploration more practical, while NASA still recommends engineering review and validation when established CEA2 workflows move to v3—particularly for mission-critical analyses or cases that depend on legacy assumptions.