Well, that greatly depends on the thinking required to produce a single LDraw model, not so much regarding how much of a "thinker" a model is, but how much thinking it must do to get a model right.
Technic models, for example, are usually expensive, given that initial placements are usually further from the right position than non-Technic models, the "bricks" ones.
A (wild!) estimate: Astra, building a Technic mechanism, around 5$...
This is one of the reasons for creating this project: making a tooling that can be used by low-end agents to iteratively build, while high-end agents, like e.g. Astra or Opus, would do the planning, vision models the inspection... etc.
Well, who knows :)
How well does this work with local models like https://github.com/receptron/laya (open source Jev) and vision capable models like Qwen 3.8 Flash Next, or even Qwen 3.8 27B on xhigh thinking?
Didn't try this with laya, but just tested it with Qwen 3.8 27B via OpenRouter.
Tried twice, one starting directly in Agent mode, and another one starting first in Plan mode and then Agent.
Conclusions: very slow in general, starting in Plan mode gives way better and faster results, but still with defects.
See: - Notes: https://github.com/anteloc/ldraw-nova-docker/blob/experiment... - Result: https://github.com/anteloc/ldraw-nova-docker/blob/experiment...