Insights 5 min read

What Two Months of SysML v2 Evaluation Taught Us About the Future of MBSE

What Two Months of SysML v2 Evaluation Taught Us About the Future of MBSE

On June 6th, the systems engineering community closed the book on what we believe is the industry's first two-month vendor evaluation of SysML v2. Five premier companies put their tools in front of a room of practitioners, answered hard questions, and, in the part that mattered most, critiqued each other's work.

SysModeler.ai was one of them. Here's what happened, and why we think this format deserves to become the norm.

Why an evaluation like this was overdue

SysML v2 is not a version bump. It is a ground-up rework of how systems models are expressed, stored, and exchanged: a formal semantic foundation, a textual notation alongside the graphical one, and a standard API and services layer that finally makes model interchange a first-class concern rather than an export-and-pray afterthought.

That last part is the promise. The open question, the one every systems engineering organization is quietly asking before it commits budget and migration effort, is whether the promise survives contact with actual tools built by actual vendors with actual differing interpretations of the spec.

You cannot answer that question with a datasheet. You cannot answer it with a forty-minute conference demo, either. You answer it by giving multiple vendors the same problem, the same amount of time, and a room full of people qualified to tell when something is being glossed over.

That is exactly what this evaluation did.

Credit where it's due

None of this happens without organizers willing to absorb two months of coordination work for no commercial upside. The INCOSE Michigan Chapter hosted and ran the evaluation, with Michael Vinarcik, ESEP-Acq, P.E., FESD and Karl Selewski leading it.

Getting five competing vendors to participate in a structured, months-long, publicly reported evaluation is a genuinely hard organizational problem. Getting them to sit in a room afterward and critique each other's models is harder. That it happened at all is a testament to how the effort was led.

Our thanks also to the other participating companies, including Dassault Systèmes, for showing up and engaging in good faith.

Our showcase

During the showcase we demonstrated SysModeler.ai's reliable AI capabilities and the matured feature set we've built to support digital engineering and digital twin workflows.

The word we care about there is reliable. AI assistance in an MBSE tool is easy to demo and hard to trust. A model is not a chat transcript; it is an artifact that downstream analysis, verification, and eventually hardware decisions depend on. Plausible-looking output that quietly violates the spec is worse than no output at all, because someone has to find it later.

Building for that standard is most of the work. It's also the part that only shows up under exactly this kind of extended, scrutinized evaluation, which is why we wanted to be in it.

You can watch our full demonstration here: https://lnkd.in/gUHcXrDX

The afternoon interoperability session was the real event

If you take one thing from this post, take this: the most valuable hours of the day were the ones where vendors anonymously critiqued each other's models.

Anonymity changed the character of the feedback. Nobody was defending a brand. The critiques were specific, technical, and occasionally uncomfortable, which is what useful critique tends to be. And because everyone in the room had spent the same two months on the same problem, the feedback landed with people who could actually act on it.

What emerged was a transparent, honest look at how these tools handle real-world model exchanges. Not the sanitized version. Not the version where the exchange is between two instances of the same tool. The version where a model built under one vendor's interpretation of the spec meets another vendor's interpretation, and you find out precisely where the seams are.

Some of those seams are spec ambiguities that the standards community should hear about. Some are implementation gaps that individual vendors, us included, need to close. Both are worth knowing, and you cannot learn either one alone.

What we'd like to see next

The industry needs more events structured exactly like this. A few things made it work:

  • Duration. Two months is long enough that you cannot special-case your way through the problem. Tools get evaluated as tools, not as demos.
  • A common problem set. Shared inputs make the comparison meaningful.
  • Anonymous peer critique. This is the piece most events skip, and it is the piece that produced the most learning.
  • Neutral organizers. An INCOSE chapter has no stake in which vendor wins. That neutrality is what made vendors willing to participate honestly.
  • Public output. Findings that stay in the room help nobody make a decision.

SysML v2 adoption is going to be decided over the next few years by organizations weighing migration cost against interoperability benefit. Evaluations like this one give them real evidence to weigh. The alternative, in which every buyer independently runs their own private bake-off, is enormously wasteful and produces results nobody else can learn from.

Closing

It was a great opportunity to meet and learn from peers who are actively shaping the future of MBSE. We came away with a clearer picture of where the standard is solid, where it is still settling, and what we need to build next.

We'd participate again without hesitation. We hope other chapters, consortia, and standards bodies pick up the format.

If you're evaluating SysML v2 tooling and want to talk through what we learned, we're happy to have that conversation.

SysML v2

SysModeler Team

Engineering & Product