# Solidcore vs OCCT: a young promise next to a thirty-year-old kernel

- Link: https://opensourcecad.nl/en/solidcore-vs-occt/
- Published: 2026-10-22T21:43:21+02:00
- Author: Marijn van den Heuvel \+ AI

_This is the English reading version of the Dutch news item [Solidcore vs OCCT: jonge belofte naast de kernel van dertig jaar](https://opensourcecad.nl/vibecoded-apps/solidcore-vs-occt/)._

OpenCASCADE (OCCT) is the kernel almost every open-source CAD application runs on.
Solidcore is the newcomer, a rebuild of the kernel that sat inside ClassCAD for 
years and was taken over by NVIDIA in 2022. It is a project by the makers of SolidCraft
itself, which still runs on the Truck kernel and will move over later. The question
underneath is simple: is this a replacement or a promise?

|  | OpenCASCADE (OCCT) and Solidcore | 
| Age | OCCT: in use since 1999. Solidcore: the rebuild began on 12 October 2026; the original kernel has run for over twenty years in ClassCAD. | 
| Shapes | Both work with analytic shapes (planes, cylinders, cones, spheres) plus NURBS. | 
| Proven in | OCCT: FreeCAD, CadQuery, Build123d, KiCad 3D, Salome. Solidcore: nothing yet; SolidCraft still runs on the Truck kernel and moves over later. | 
| Status | OCCT: mature, decades in production. Solidcore: 197 of 261 parts ported; by the project’s own account nothing yet works end to end. | 
| Stewardship | OCCT: Open Cascade SAS, part of Capgemini since 2014; development follows its paying customers, who build geometry exchange tools and 3D viewers. Solidcore: an independent project by one developer, the same maker as SolidCraft. | 
| Speed | OCCT: known for slow fillets, booleans and STEP import on heavy models. Solidcore: a promise of better performance; not a single measurement published yet. |

What a kernel does, you know if you work with open-source CAD: it is the computing
heart that turns your sketch into a real solid, that works out booleans, applies
fillets and writes your model to STEP. OCCT has done that since 1999, almost every
open tool leans on it, and there is no second open-source kernel that has been in
production that long. The downside sits in the stewardship. Behind the kernel stands
Open Cascade SAS, part of Capgemini since 2014, and you notice that in where the
development goes. The paying customers of that company do not build CAD programs
with the kernel: they are parties making tools for geometric exchange and 3D viewers,
with visualisation and data integration as their trade. The roadmap follows those
customers. Version 8.0.0 took thirteen months and brought mainly a new topology 
representation, faster computational techniques and cleanup work; the next version
is about the visualisation layer and data exchange. Development is not standing 
still, but the priority lies elsewhere: FreeCAD, the largest open-source CAD program
on this kernel, is not a paying customer for Open Cascade and therefore not a priority.
Anyone who needs fillets, booleans and STEP import notices that.

Solidcore comes from a different corner. SMLib is the kernel that earned its stripes
in ClassCAD, was taken over by NVIDIA in May 2022 and appeared in October 2026 as
the open-source part of USD BRep. One developer is now rebuilding it, with a hard
measuring stick: no crashes on unexpected input, the same outcome for the same input,
and every computation has to match the original down to the last decimal. He tests
that by running both kernels side by side on the same input.

The honesty of that project is striking. The progress page says exactly where it
stands, in its own words: nothing works end to end yet, there are no primitives,
booleans or fillets. What you do see is pace: the rebuild began on 12 October and
a week later 197 of the 261 parts had been ported, about seventy percent of the 
work.

Why two projects look at it that way: SolidCraft’s own kernel planning says that
slow fillets, booleans and STEP import are the real pain, and that Solidcore has
to become the solution. And in the discussion about CADRS the maker calls OCCT somewhat
slow, and says a switch is in principle the conversion of one module. He is building
a kernel of his own alongside it, to compute on the graphics card.

**My verdict:** I am sceptical about OCCT, and that is not about the age of the 
code but about who earns from it. The company behind it lives on customers who use
the kernel for geometry translation programs and viewers, not to build CAD programs.
Development is not standing still, but where the priority lies follows from who 
pays, and you feel that in the fillets, booleans and STEP imports of the programs
running on it. At the same time it is the only open-source kernel that can carry
a model from sketch to production, so there is no alternative today. For now Solidcore
is not a replacement but a promise with a hard measuring stick under it. Compare
the performance only once the thing works end to end, and do not be tempted by the
percentage on that progress page: ported parts are not a working fillet yet.

The occasion for this comparison is in the news item [CADRS considers another kernel](https://opensourcecad.nl/en/cadrs-considers-another-kernel-from-opencascade-to-solidcore/).
What a kernel does and why the choice is so decisive is on [NURBS & modeling kernels](https://opensourcecad.nl/en/nurbs-modeling-kernels/).**
Keep an eye on this site for updates and tests**, I keep following both kernels 
and will publish the first measurements here as soon as they exist.