Benchmark
2D Airfoil · External aero
Engineering question: Can a surrogate reproduce lift, drag, and surface pressure across angles of attack on a 2D airfoil?
One airfoil, many angles of attack. Lift and drag for performance. Surface pressure for how the wing loads up. Classic teardrop cross section. The smallest step into external aero before a full wing or vehicle.
Angle of attack sweeps the operating point
- Industry
- Aerospace / general aero
- Domain
- Fluid dynamics 2D
- Dataset
- 2D airfoil
- Task
- Surface pressure + forces
- Solver
- OpenFOAM
- Input
- Angle of attack
- Train
- 15 cases
- Holdout
- 5 test cases
Scope and limits
Comparability
2D external aero, not comparable with 3D vehicle or structural benchmarks. Mesh and Reynolds number are fixed within the open training corpus.
Ground truth
CFD from Rescale AI Physics training exports (OpenFOAM). Holdout cases use simulation truth for Cl, Cd, and nodal Cp.
Submissions
Published scores from models trained on the open 2D airfoil split and evaluated on shared holdout cases. Check rows to compare; expand for per-AoA predictions and Cp field metrics.
Select submissions to compare
| Compare | Model | Submitted | Train time | Cl acc. | Cd acc. | Cp R² | Details |
|---|---|---|---|---|---|---|---|
| Two-Stream Transformer7485a | Jun 17, 2026 | 1m 50s | 97.13% | 98.83% | 0.9911 | ||
| DoMINO4dfb6 | Jun 17, 2026 | 32m 50s | 99.33% | 99.79% | 0.9387 | ||
| MeshGraphNet5d7f7 | Jun 17, 2026 | 15m 8s | 89.97% | 97.33% | 0.9831 | ||
| Transolver68434 | Jun 24, 2026 | 9m 17s | 92.72% | 98.56% | 0.8728 |
Ready to try this out? Download the training split, or submit if you already trained on the shared protocol. Check the submission guide →