Spotlight: XGRIDS Lixel L3, and a 5 mm walk into the model
XGRIDS launched Lixel L3 at INTERGEO with validated 5 mm relative accuracy. What scan-to-BIM teams should test before they trust a walk.
Monday’s unused leftover was a walk-scan into MAGNET. The INTERGEO handheld that still has a BIM test left is XGRIDS Lixel L3.
That is not a new authoring tool. Firms already walk a floor plate, then argue whether the cloud is good enough to model from. XGRIDS’s claim, announced 15 September and repeated by Lidar News on 16 September, is that a handheld SLAM unit can hold a documented 5 millimetre relative accuracy and still export locally into the BIM path the firm already runs.
What actually changed
Lixel L3 is the flagship XGRIDS handheld, not a CDE. The launch note puts LiDAR, a 1/1.3-inch CMOS 4K HDR colour system, integrated RTK and a 4.3-inch display in one kit. Dealers list two LiDAR builds on the same 120-metre range: 16-channel at 320,000 points per second, and 32-channel at 640,000. Weight with battery is about 2.5 kilograms. IP54. About 100 minutes on a charge. XGRIDS prints around 100 square metres in 10 minutes for a high-accuracy walk plus a photorealistic model.
The number that matters is relative, not absolute. The INTERGEO note lists 5 millimetres relative accuracy and about 2.5 centimetres real-time absolute accuracy. That split is the whole story. Relative tells you whether two extracted points stay the right distance apart. Absolute tells you whether the cloud sits on the project grid. A walk that is 5 millimetres internally and 25 millimetres off the control is still a failed as-built.
XGRIDS published a point-cloud accuracy white paper behind the headline. Seven L3 units, four 16-channel and three 32-channel, ran 20 independent indoor scans with five operators. The site was about 507 square metres of office: long corridors, rooms and a stairwell. A Leica TS60 set 13 check targets. Targets were not used as GCPs. Processing ran through Lixel Studio 4.1. Across 1,524 point-pair distances from 5.4 to 44.2 metres, the pooled P68 error was 4.2 millimetres. That supports the 5 millimetre at 1σ typical specification under those conditions. It is not a guarantee on your plant room.
The software path is local. Process the cloud in LixelStudio. Rebuild the splat in Lixel CyberColor if you need a photorealistic walkthrough. Outputs on the launch note are a true-colour point cloud, a mesh and a 3D Gaussian splat in XGRIDS’s LCC2 format. Map fusion can align multiple walks without preset control points. Privacy tools can blur faces and plates. The BIM sentence from XGRIDS is that the data can go straight into the workflow after a local export. That is a file, not a native Revit session. You still lock a coordinate system, index the cloud and compare the finished model back to the source. Treat the splat as a visualisation of the same walk.
The INTERGEO note has no public list price. Confirm a dealer kit, LixelStudio seats and lead time before you write it into a method.
What a BIM or VDC lead should do
Do not rewrite the scan-to-BIM standard because the relative figure got smaller. The rules in our scan-to-BIM note still hold: say what the cloud is for, set a tolerance, lock a coordinate system, and keep a read-only source.
If the firm already loses a day sending a crew back for a corridor the static scanner never reached, this is worth a side-by-side test. Pick one non-sensitive floor you already modeled from a terrestrial cloud. Walk it with L3. Process it locally. Then ask four questions: does the on-device display show coverage before the crew leaves; do extracted distances hold against your own tape or total station, not just the vendor white paper; does the exported cloud sit on the control you already trust; and can a modeler pick the same door, grille or hanger they picked from the static scan.
If the answers hold, you have a shorter path from walk to a usable cloud. If a 15 millimetre hanger disappears, or the cloud is 25 millimetres off the grid, keep the tripod. Lixel L3 is a documented 5 millimetre walk. It is not proof that the wall in the model is the wall on site.
Start with XGRIDS’s 15 September note, the Lidar News restatement, and the product page.
Frequently asked questions
What did XGRIDS announce?
On 15 September 2026 at INTERGEO in Munich, XGRIDS introduced Lixel L3, a handheld spatial scanner that combines LiDAR, a 4K HDR colour system, built-in RTK and a 4.3-inch on-device display. The same note says one walk can produce a true-colour point cloud, a mesh and a photorealistic 3D Gaussian splat. Processing can stay local.
Is 5 mm the accuracy of a project cloud?
No. That is XGRIDS’s typical post-processed relative accuracy at 1σ, from a documented indoor test. Real-time absolute accuracy in the launch note is about 2.5 centimetres. Relative distance is not a closed coordinate system. Keep checkpoints, a CRS statement and a model-to-cloud check.
Does this replace a terrestrial laser scanner?
Not for millimetre plant-room work. Five millimetres of relative SLAM is tighter than centimetre-class walk-scans, and still a long way from a static object-accuracy figure such as FARO’s 0.9 mm on Focus i.1. Test a packed corridor before you park the tripod.
What lands in the BIM workflow?
XGRIDS says crews can capture, process and export locally, then send the data into the existing BIM path. The useful file is still a point cloud you index and georeference. The 3D Gaussian splat is a visualisation of the same walk, not a coordinated model.