Shopify 3D model size limits: what the rules are and how to fix them

Shopify enforces two hard limits on 3D product models uploaded to a product page: files must be under 15 MB, and models should stay under 100,000 trianglesfor reliable 3D viewer and AR Quick Look performance. Both limits are documented in Shopify's official guidelines, and exceeding them causes the 3D experience to fail — quietly, and in ways that cost you sales.

Why 15 MB matters for load time

Your Shopify product page 3D viewer downloads the full GLB before it renders anything — there is no progressive streaming. On a typical 4G connection (~10 Mbps), a 15 MB file takes around 12 seconds to arrive. On congested or slow mobile networks, that climbs to 20–30 seconds.

Research on mobile e-commerce consistently shows that most shoppers abandon a page if interactive content doesn't appear within 3–5 seconds. When a product model takes 12 seconds, the 3D viewer has the opposite of its intended effect — it signals that the site is slow and unreliable. A model comfortably under 5 MB — ideally 2–4 MB — loads in under 2 seconds even on a poor connection and keeps shoppers engaged.

Why 100,000 triangles matters for AR

AR Quick Look on iOS and Google Scene Viewer on Android render your model at 60 fps in a live camera feed, on mobile hardware. A model with 480,000 triangles will drop frames, stutter, or fail to launch entirely on mid-range devices — the phones that the majority of your customers use. Shopify's 100k triangle ceiling is calibrated to what mobile AR hardware handles smoothly and reliably, not just on high-end flagships but on 2–3 year-old mid-range devices.

Exceeding the limit doesn't guarantee a crash, but it makes AR unreliable across the fragmented device landscape you cannot control. The result: a "View in your space" button that works in your QA environment on a new iPhone and fails on your customers' actual phones.

What actually makes a GLB over Shopify's limits

Almost all overweight Shopify models fail the same three ways:

• Uncompressed textures. A single PNG at 4096×4096 pixels takes ~48 MB of GPU memory when expanded and several megabytes on disk. Four of those in one model and you are already 10× over budget on texture data alone. Converting to KTX2 (Basis Universal) cuts texture data 70–85% with no visible quality loss.
• No geometry compression.Models exported directly from CAD software, Blender, or photogrammetry tools ship raw mesh data. Draco geometry compression — supported natively by Shopify's viewer — typically cuts geometry size by 70–80%.
• Too many triangles. Products modeled for rendering (not real-time use) routinely arrive with 500k to 2M+ triangles. Web viewers and AR engines need 30–100k for most product types. Selective decimation removes the excess while protecting silhouettes and material seams.

How to diagnose your model

Before fixing anything, you need to know exactly where your model is over budget. MeshGrade's free Shopify analyzer runs in your browser — no upload, no sign-up — and reports:

• File size vs. the 15 MB ceiling
• Triangle and vertex counts vs. the 100k target
• Every texture with its resolution, format, and GPU memory cost
• Estimated draw calls
• A 0–100 grade benchmarked against Shopify's requirements

The report is a prioritized list of what to fix, ordered by impact. You will know whether the issue is textures, geometry, or both — and roughly how much each fix will reduce the file.

How to fix it: three steps in order

Step 1 — Compress textures to KTX2 (biggest single impact)

Textures are almost always the largest contributor to GLB file size. MeshGrade's texture analyzer flags every uncompressed or oversized texture. Converting to KTX2 via Basis Universal is the single highest-impact change for most models. Textures at 4096×4096 PNG can often be resized to 2048×2048 KTX2 with no perceptible quality difference on a product viewer.

Step 2 — Apply Draco geometry compression

Draco compresses mesh geometry at encode time and is decoded transparently by Shopify's viewer. For most product models, Draco reduces geometry data by 70–80%. Run your GLB through the file size analyzer to see the before-and-after impact on your specific model.

Step 3 — Decimate if you're over 100k triangles

For Shopify product viewers, 50–80k triangles is the reliable target. More is rarely visible on screen at typical viewing distances, but it directly affects AR stability on phones. Selective decimation protects silhouette edges, UV seams, and material boundaries while removing geometry the renderer never needed. Check your polygon count →

Before and after: furniture model

A furniture model graded on MeshGrade came in at 22 MB and 480,000 triangles — grade D. AR Quick Look consistently failed to launch on every iPhone used for testing.

After optimization:

• File size: 3.4 MB (85% reduction)
• Triangles: 70k (85% fewer)
• Grade: A
• AR Quick Look: launches instantly, reliable across device types

The silhouette and material fidelity were identical to the original. The only difference a customer would notice is that AR now works instead of silently failing.

The cost of not fixing it

The business case is direct. AR product experiences consistently show higher conversion rates compared to non-AR sessions — but that conversion premium only materializes if AR actually launches. A model over Shopify's limits guarantees that most customers on mid-range phones never experience it in AR. Every heavy model is a quiet conversion leak.

Next step

Grade your product model free → — no sign-up, runs in your browser. You'll know within 30 seconds whether you're inside Shopify's limits and what to fix if you're not.