Rapid Prototyping
Rapid prototyping builds a part layer by layer directly from a CAD model, with no tooling and no minimum order quantity.
- Part
- Forming
- Variants
- 3
- Revised
- 2026-08-11
At a glance
- Family
- Additive Manufacturing
- Typical tolerances
- Process-dependent: ±0.004 in (±0.1 mm) for SLA, MSLA, PolyJet, and metal powder bed fusion; ±0.008–0.012 in (±0.2–0.3 mm) for MJF and SLS; ±0.020 in (±0.5 mm) for desktop FDM. Machined features are tighter than any of these.
- Surface finish
- Ra 20 µin (0.5 µm) on PolyJet and SLA up-facing surfaces through Ra 1,000 µin (25 µm) on coarse FDM and Ra 1,400 µin (35 µm) on EBM. See the [surface finish chart](/charts/surface-finish-chart) for comparison against machined and cast finishes.
- Typical volumes
- 1–1,000 parts; unit cost is essentially flat with quantity
- Lead time
- 1–5 business days for plastic parts; 5–15 business days for metal, where heat treatment and machining dominate. No tooling lead time in any case.
- Materials
- Plastic, Metal, Composite
What it is
Rapid prototyping is the practice of producing a physical part directly from a CAD model with no tooling and no minimum order quantity — in practice, additive manufacturing plus the low-volume processes that share its economics, such as CNC machining and vacuum casting. The defining characteristic is that unit cost is nearly flat with quantity: the first part costs what the hundredth does, because there is no mold to amortize.
Lead time is the reason it exists. A part that would take 4–8 weeks to tool can be in hand in 1–5 business days, which changes how many design iterations a program can afford.
The trade-offs are real. Accuracy ranges from ±0.020 in (±0.5 mm) on desktop FDM to ±0.004 in (±0.1 mm) on SLA and metal powder bed fusion; material properties are process-specific rather than chosen freely; and above roughly 1,000–10,000 parts, molded and cast processes win decisively on unit cost.
How it works
Every additive process follows the same five steps; what differs is how each layer is created.
- Prepare the model. Export a watertight mesh (STL, 3MF, or STEP for machining) and check for inverted normals and non-manifold edges. Wall thicknesses that were legal in CAD may be below the chosen process's minimum.
- Orient and support. Orientation sets accuracy, surface finish, strength direction, and cost simultaneously — it is the single most consequential decision in the workflow. Powder-bed processes (SLS, MJF, binder jetting) need no supports; vat, extrusion, and metal powder-bed processes do.
- Slice. Layer heights range from 0.0006 in (14 µm) on material jetting to 0.016 in (0.4 mm) on coarse FDM. Build time scales roughly inversely with layer height.
- Build. Point-scanning processes (SLA, SLS, DMLS) take time proportional to cross-sectional area; whole-layer processes (MSLA, MJF, binder jetting) take time proportional to height only, which is why they scale better with part count.
- Post-process. Support removal, washing, post-curing, depowdering, stress relief, or sintering, depending on the process. This step is routinely underestimated and is often the majority of the lead time for metal parts.
Which process for which job?
| Process | Layer height | Typical tolerance | Materials | Best for |
|---|---|---|---|---|
| SLA | 0.001–0.004 in (25–100 µm) | ±0.006 in (±0.15 mm) | Photopolymer | Smoothest surface, fine detail, master patterns |
| MSLA / DLP | 0.001–0.004 in (25–100 µm) | ±0.004 in (±0.1 mm) | Photopolymer | Many small parts per plate; dental, jewelry |
| FDM | 0.002–0.016 in (0.05–0.4 mm) | ±0.020 in (±0.5 mm) | Thermoplastic | Cheapest, largest, real engineering resins |
| SLS | 0.004 in (100 µm) | ±0.012 in (±0.3 mm) | Nylon, TPU | Functional plastic parts, no supports |
| MJF | 0.0031 in (80 µm) | ±0.008 in (±0.2 mm) | Nylon | Functional nylon at 10–10,000 pieces |
| PolyJet | 0.0006–0.0013 in (14–32 µm) | ±0.004 in (±0.1 mm) | Photopolymer | Multi-material, full color, overmold models |
| DMLS / SLM | 0.0008–0.0024 in (20–60 µm) | ±0.004 in (±0.1 mm) | Metal alloys | Metal geometry that cannot be machined |
| CNC machining | n/a | ±0.005 in (±0.13 mm) | Metal, plastic, wood | True material properties, tight tolerance |
Design guidelines
Choose the process before you finish the CAD
Minimum wall, minimum hole, and support requirements differ by a factor of three or more across processes. Designing a part to "3D printing" in the abstract produces geometry that is expensive in every one of them.
Wall thickness by process
0.020 in (0.5 mm) for SLA and MJF, 0.030 in (0.8 mm) for SLS, 0.032 in (0.8 mm) for FDM, 0.040 in (1 mm) for metal powder bed fusion. Below these, parts either fail during the build or distort during cooling and cleaning.
Orientation drives everything
Tolerance, surface finish, strength direction, support volume, and price all follow from how the part sits in the machine. State the critical surface on the drawing rather than leaving orientation to the operator.
Do not print threads under M6
Print a pilot hole and tap it, or design for a heat-set or press-in insert. Printed threads in any additive process are dimensionally poor and strip early. The tap drill chart gives pilot sizes and the thread size chart the corresponding major diameters.
Add stock where tolerance matters
No additive process reliably holds a bearing fit. Print the feature 0.020–0.040 in (0.5–1.0 mm) undersize or oversize as appropriate and machine or ream it. See the ISO 286 fits and tolerances chart for the fit classes and clearances involved.
Design escape routes for powder and resin
Every enclosed volume in a powder or vat process must be openable. Powder needs 0.20 in (5 mm) ports; resin needs 0.12–0.16 in (3–4 mm) drains, two per cavity.
Know when to leave
Above roughly 1,000 parts for small plastic components, injection molding is cheaper per part despite the tooling. For 50–500 cosmetic parts, an SLA master plus vacuum casting usually beats printing them all.
| Feature | Recommended | Limit | Why |
|---|---|---|---|
| Wall thickness | 0.040 in (1.0 mm) | 0.020 in (0.5 mm) process-dependent | Build and cleaning survivability |
| Hole diameter | 0.12 in (3 mm) | 0.020–0.060 in (0.5–1.5 mm) | Small holes close in every process |
| Threaded feature | Insert or tapped | No printed thread under M6 | Layer structure strips under load |
| Machining stock | 0.040 in (1.0 mm) | 0.020 in (0.5 mm) | Additive cannot hold a bearing fit |
| Escape port | 0.20 in (5 mm) | 0.12 in (3 mm) | Powder and resin must exit |
| Overhang, supported processes | 45° from vertical | Support required | Each layer needs the one below |
Cost drivers
Rapid prototyping has no tooling cost, so the entire bill is machine time, material, and post-processing labor. Machine time is dominated by build height in every process, and additionally by cross-sectional area in the point-scanning ones. Material cost tracks part volume, including support and, in powder processes, the un-fused powder that cannot be recycled.
Post-processing labor is the item that surprises people. Support removal, sanding, dyeing, depowdering, stress relief, and machining are hand operations that scale linearly with quantity and often exceed the machine cost. For metal parts, post-processing routinely costs more than the print.
Volume breakpoints worth memorizing: below 10 parts, almost any additive process is cheaper than tooling. From 50 to 500 cosmetic plastic parts, an SLA or MSLA master plus vacuum casting typically wins. From 500 to 10,000 functional nylon parts, MJF or SLS is competitive. Above 1,000–10,000 depending on part size, injection molding wins on unit cost and does not look back.
- Reduce height before anything else; it is the strongest cost lever in every process.
- Hollow, rib, or lattice solid volume rather than shrinking the envelope.
- Batch parts into one build; machine setup and cooldown cycles are shared.
- Loosen tolerances you do not need and call out only the features you do.
- Match the process to the quantity honestly — printing 2,000 parts because the first 20 printed well is the most common and most expensive mistake in the category.
Variants
3 named
Stereolithography (SLA)
Selective Laser Sintering (SLS)
Direct Metal Laser Sintering (DMLS)
Questions
6 questionsWhich rapid prototyping process should I use?
For the smoothest cosmetic model, SLA. For functional plastic parts, SLS or MJF. For the cheapest large part or a real engineering thermoplastic, FDM. For multi-material or full-color models, PolyJet. For metal, DMLS or SLM. For tight tolerance and true material properties, CNC machining.
How fast is rapid prototyping?
1–5 business days for most plastic parts and 5–15 for metal, where stress relief and machining dominate. The comparison that matters is against tooling: an injection mold takes 4–8 weeks, so prototyping buys iteration cycles rather than just parts.
At what quantity should I stop prototyping and buy tooling?
For small plastic parts, injection molding generally wins somewhere between 1,000 and 10,000 pieces depending on part size and complexity. Between about 50 and 500 cosmetic parts, an SLA master plus urethane vacuum casting usually beats both printing and tooling.
How accurate are 3D printed parts?
It varies by an order of magnitude across processes: ±0.004 in (±0.1 mm) for SLA, MSLA, PolyJet, and metal powder bed fusion, ±0.008–0.012 in (±0.2–0.3 mm) for MJF and SLS, and ±0.020 in (±0.5 mm) for desktop FDM. No additive process holds a bearing fit — leave stock and machine the feature.
Can prototype parts be used as end-use parts?
Some can. SLS and MJF nylon and metal powder bed fusion parts are used in production regularly. Photopolymer parts from SLA, MSLA, and PolyJet generally cannot, because they embrittle under UV and creep under sustained load.
Why does part orientation matter so much?
Orientation simultaneously determines build height (cost), which surfaces carry supports (finish), the direction of the weak axis (strength), and the accuracy of specific features. It is the single most consequential choice in the workflow, which is why the critical surface should be called out on the drawing.