We Analyzed 246,000 3D Models. Here's What We Found.

ModelDirectory.org aggregates 3D models from four major platforms: Printables, Thingiverse, MyMiniFactory, and Sketchfab. We currently index 246,181 published models from 31,361 creators. That dataset represents nearly half a billion downloads.

I ran the numbers on the full database. What follows is a snapshot of the 3D model ecosystem in mid-2026 — file formats, platform dynamics, category distribution, licensing, and download patterns. Every number comes directly from our database, not surveys or estimates.

Platform Breakdown: Printables Dominates

The distribution across platforms is not even close. Printables accounts for more than three quarters of all indexed models, and its share of total downloads is even more extreme.

Several things stand out. Printables has 3.4x more models than Thingiverse and 77x more recorded downloads. But the download disparity is partly because Thingiverse's API doesn't expose download counts as reliably — this is a data coverage issue, not necessarily a real gap of that magnitude.

MyMiniFactory and Sketchfab together make up less than 1.5% of indexed models. Both platforms serve different niches — MyMiniFactory leans toward miniatures and tabletop gaming, Sketchfab toward visualization and non-printable 3D — but their share in the printable model ecosystem is small.

File Formats: STL Still Rules, But 3MF Is Growing

The format question gets debated endlessly in 3D printing communities. The data tells a clear story.

STL at 94.6% is overwhelming. Despite years of advocacy for 3MF (multi-color support, print settings, metadata), it has reached just 2.4% of published models. STEP — the format mechanical engineers prefer for its parametric editability — sits at 1.0%.

The remaining formats combined (OBJ, BLEND, GLB, FBX, and others) account for under 1%. For practical purposes, the 3D printing ecosystem runs on STL. If you're building tools or viewers for this space, STL support is non-negotiable. Everything else is a nice-to-have.

What People Are Printing: Categories

We use AI classification to tag every model into a primary category. The top categories show a community that's practical, creative, and a little playful.

Organisation tops the chart — bins, trays, holders, cable management, desktop organizers. It's the unsexy backbone of consumer 3D printing. People buy printers to print functional objects that make their space tidier.

Art at 21% includes figurines, sculptures, decorative pieces, and increasingly, articulated print-in-place toys. The most downloaded model in our database — Honeycomb Storage Wall by RostaP with 271,060 downloads — sits right at the intersection of art and organization.

The Most Downloaded Models

The top 10 most-downloaded models reveal what actually gets printed at scale.

Three themes dominate the top 10: functional prints (storage wall, printbed scraper, cable organizer), articulated toys (octopus, dragon, shark, jointed figure), and decorative objects (moon lamp, spiral planter). McGybeer appears twice — the Cute Mini Octopus and the Articulated Shark — making them arguably the most successful individual creator in the ecosystem by download volume.

Licensing: The CC-BY-NC-SA Default

Creative Commons licenses dominate the 3D printing ecosystem. Here's how creators choose to license their work.

CC BY-NC-SA is the most popular license by a wide margin. Creators generally want attribution and don't want their designs sold commercially — but they're fine with remixing as long as derivatives carry the same terms. Only 2.8% of models use CC0 (public domain), and the restrictive CC BY-NC-ND (no derivatives allowed) accounts for about 4.8%.

For anyone looking to sell printed models commercially, the practical filter is: 19% of the ecosystem uses CC BY or CC BY-SA licenses, which permit commercial use. The rest either restricts commercial use or doesn't specify.

AI-Generated Models: Still a Fraction

There's been a lot of discussion about AI-generated 3D models disrupting the community. The data suggests we're still in very early days.

Under 2% of published models carry the AI-generated flag. This number is almost certainly an undercount — platforms only recently started requiring AI disclosure, and many AI-generated models were uploaded before the flag existed. Still, even with undercounting, AI models are not yet a significant portion of the ecosystem by volume.

Remixes and Derivatives

The remix culture in 3D printing is real but modest. 6,755 models (2.7%) are flagged as derivatives of other models. This includes remixes, mashups, and modifications of existing designs.

The relatively low remix rate might surprise people who think of 3D printing as inherently remix-friendly. Part of the explanation is data: not all platforms track derivative relationships, and many remixes aren't formally linked to their parent models.

The Average Model

If you averaged every model in the database, here's what you'd get:

• Format: STL (94.6% probability)
• Platform: Printables (76.3%)
• Category: Organisation or Art (42.8% combined)
• License: CC BY-NC-SA (36.1%)
• Downloads: 1,964 (mean) — though median is much lower
• AI-generated: No (98.2%)
• Is a remix: No (97.3%)
• Cost: Free (effectively 100% — zero paid models in our current dataset)

What This Data Doesn't Show

A few honest caveats. Our index covers four platforms but not all of them equally. Printables has the deepest coverage because their API is the most complete. Thingiverse download counts are underrepresented. MyMiniFactory and Sketchfab are partially indexed. Cults3D — a significant marketplace — is not yet indexed at all.

The numbers here are a floor, not a ceiling. The actual 3D model ecosystem is larger than what any single directory captures.

Where This Is Going

We'll be updating these numbers monthly. As our index grows and we add more platforms, the picture will get more complete. If you want to explore the models behind these statistics, every model has its own page on ModelDirectory.org — with the original creator credited, the source platform linked, and a SHA-256 fingerprint for provenance.

If you're a researcher, journalist, or creator who wants to dig into this data — keep reading.