Not every drawing review is the same. Some checks are routine, like confirming the title block is filled out, verifying that approved materials are listed, and making sure general notes meet compliance. Other checks are harder, requiring you to read the geometry, interpret design intent, and determine whether a main datum feature flag is actually placed on a measurable surface.
Today we are launching the bananaz Pro Model, purpose-built for the complex, multi-step engineering tasks that demand deeper reasoning. It runs alongside our existing Light Model, which is built for fast checks and high-volume drawing reviews. This post covers what each model does, when to use them, and why splitting AI drawing review into two tiers changes how mechanical engineering teams scale quality.
The Trade-Off Between Speed and Depth
Drawing review has always involved a trade-off between speed and depth. A fast review catches the obvious errors like missing dimensions or misaligned title blocks, but it cannot reason about GD&T intent or judge whether a datum scheme is functionally correct. A deep review can reason through a tolerance stackup or evaluate datum feature flags, but running that level of analysis on every drawing across a 200-part assembly is overkill that burns time and cost on routine checks.
Mechanical engineering teams need both a fast first-pass for high-volume compliance work and a deeper second-pass for the parts of the drawing where senior engineering judgment actually matters.
Meet the Two Models
Light Model: Fast Checks at Scale
The Light Model is built for the high-volume, straightforward rule-based work that fills your daily review queue.
Use Light for:
- Title block validation
- Approved material checks
- General notes compliance
- Standard drawing format reviews
- PLM and drawing BOM comparison
Light moves fast, flagging the issues you would catch in a five-minute sanity check, except it does it quicker.
Pro Model: Deep Reasoning for Complex Tasks
The Pro Model is built for the work that used to require a senior engineer in the room.
Use Pro for:
- GD&T validation and datum scheme review
- Determining whether datum feature flags are placed on measurable surfaces
- Manufacturability judgment calls
- Regulatory standard checks (ASME, ISO, and similar)
- Anything that requires interpreting design intent across multiple views
Pro is slower than Light because it is doing more, reading geometry, cross-referencing callouts, and reasoning about how features relate to each other across the full drawing.
Ready to see Light and Pro in action on your own drawings? Book a demo with one of our AI consultants
