Really exciting stuff is happening literally every month, because underlying models are getting better and better. When I started it was pretty basic: “make a cube with a hole through it”. Now it’s at the point “make a raspberry pi 4 case” and the agent searches, builds, verifies…

What surprised me in this process is how little meaning AI benchmarks have. Pareto frontier for my use case looks completely different than any other benchmark portrays.

And this kinda makes sense. What is cheaper few KB of disk space or internet bandwidth?

Comparing 4.8 Opus with Fable 5.1

Recently we also got European funding and the project got some traction. Very exciting times ahead.

https://grandpacad.com

> Using a single LLM judge for scoring. Every score here comes from one model, GPT-5.6 Luna, looking at one image at a time. I didn’t do much alignment and didn’t check how often it agrees with itself on a re-run.

Having used a similar setup (with previous gen LLMs) to evaluate the 3D models that my product[0] generates, it turned out there was no correlation at all. LLM judgments were very much random and I assume judging SVGs is not that far from judging 3D models. I guess I have to re-test this with current gen.

[0]: https://grandpacad.com

Edit: Surprisingly very good results with 3.0 flash with high thinking.

Cost: $0.06

Duration: 3.22 min

Code Errors: 1.3 per attempts (meaning on average it had to retry 1.3 times)

Adherence was on par with 3.5 flash Low thinking

I was benchmarking using a soon to be released new version of my AI CAD modeling software[0]. It's basically an agent that has access to tools that can execute build123d scripts, get sculpted models, blender to combine sculpts + parametric models, tools to inspect the model (visually and with code), search datasheets, ...

I tried what you recommend a while ago (asking an AI to evaluate using different angles) and the AI evaluations were extremely bad - barely any correlation to what I scored. Things have gotten better, but I don't trust it enough yet.

Here is how I score adherence (and how AI did as well, but I tried methods where it would just give back a boolean "pass" or not):

    <0.2 → Poor – Misses core intent; largely irrelevant or incorrect.
    <0.4 → Weak – Partially relevant; significant omissions or errors.
    <0.6 → Fair – Covers main points but lacks completeness or precision.
    <0.8 → Good – Mostly accurate; minor gaps or deviations.
    <=1.0 → Excellent – Fully aligned; precise, comprehensive, and faithful to intent.

    dragon-bottle-stopper
    editing-param-mid-conv
    editing-parametric-enclosure
    editing-swap-material-param
    editing-text-edit-cube
    multi-turn-bird-house
    multi-turn-dice-tower
    multi-turn-modular-planter
    multi-turn-phone-stand
    multi-turn-shelf
    one-shot-bookend
    one-shot-cable-clip
    one-shot-chess-queen
    one-shot-coaster
    one-shot-coffee-cup
    one-shot-dog-tag
    one-shot-dragon-figurine
    one-shot-hex-bracket
    one-shot-keychain-fob
    one-shot-low-poly-tree
    one-shot-pegboard-hook
    one-shot-pi4-case
    one-shot-threaded-jar

Here are the results compared to Gemini 3.5 Flash:

    Model + config          CodeErr/gen   Cost/gen   Median time   Quality
    gemini-3.5-flash, low      0.71        $0.18        68s       baseline
    GLM 5.2, reasoning high    0.61        $0.18       289s         -6.0%
    GLM 5.2, reasoning off     1.52        $0.10       126s        -13.6%

Edit: I recently ran evals with Kimi 2.7 and MiniMax-M3 and this is clearly open source SOTA model, by far.

Article ends with this

- Models are very jagged (might excel in one type of 3d model, but not another)

- Gemini models are the least jagged in my experience and have the best image understanding

- Gemini models are also the most creative (which may be undesirable if you want precise CAD part)

- Overall this benchmark doesn't prove much because one 3d model (and one attempt) is just not enough. I am usually testing on at least a dozen models each generated 3 times, but should really do much more, but it's too pricey for a solo dev.

Still, thanks for publishing this. Will be definitely run flash 3.5 soon to see how it performs.