The Dawn of Vibe Modelling: Claude Opus Nails Simulation Benchmark

Published on July 7, 2025

I am officially declaring the age of 'vibe modelling' to be moving full steam ahead. You might wonder what on earth I mean by that; allow me to explain.

For a while now, I've been putting various large language models through their paces with a rather specific simulation benchmarking test. The goal is to solve a hydrogen production optimisation problem; not a trivial task, by any means. It requires a solid grasp of logic, simulation principles, and of course, the ability to write clean, functional code. The full benchmark sheet, for those who fancy a look, is available for perusal here.

Recently, I gave the challenge to Claude 4 Opus. The result, I must say, was nothing short of outstanding.

This is what I mean by 'vibe modelling'. It's the ability to use these powerful AI tools to get remarkably close to a precise, mathematically derivable answer, without necessarily needing to craft every single line of code with painstaking precision yourself. The AI captures the 'vibe' of the problem and runs with it.

Claude's code was clean, and refreshingly, not overly verbose. It adopted a proper object-oriented structure, which speaks volumes about its approach to software craftsmanship. This wasn't just a script; it was a well-thought-out piece of software.

More impressively, the answer it produced was tantalisingly close to the one you'd derive mathematically. To top it all off, it even gave me a beautiful visualisation at the end to peer at the results with. It's one thing to get a number, it's quite another to be presented with a chart that explains it.

This is a significant step forward. It suggests a future where technical professionals can rapidly prototype and solve complex problems, focusing more on the system's logic and less on the boilerplate code. The barrier to entry for high-level simulation is, it seems, getting lower.

Of course, this is just one test. But the success here has certainly given me food for thought. I think it might be time to increase the complexity of the problem and see how far this new age of vibe modelling can truly take us.