freestile(nc, "nc_mvt.pmtiles", layer_name = "counties", tile_format = "mvt")freestiler defaults to the MapLibre Tiles (MLT) format, a columnar binary tile encoding announced in January 2026 by the MapLibre organization. This article explains what MLT is, how it compares to MVT, and when you might want to use one vs the other.
Vector tiles in brief
Vector tiles divide geographic data into a grid of square tiles at multiple zoom levels. Unlike raster tiles (pre-rendered images), vector tiles store the actual geometry and attribute data, so the client can style, interact with, and dynamically render the data. The standard container for vector tilesets is PMTiles - a single-file archive you can serve from any static file host.
MVT: the standard
Mapbox Vector Tiles (MVT) has been the dominant vector tile encoding since Mapbox introduced it in 2014. It uses Protocol Buffers to serialize geometry and attributes, and it’s supported by every major mapping library: MapLibre GL, Mapbox GL, Leaflet, deck.gl, and more. When you need maximum compatibility, MVT is the safe choice.
MLT: what’s different
Where MVT stores data row-by-row (one feature at a time), MLT organizes data column-by-column - all x-coordinates together, all y-coordinates together, all values of a given attribute together. This columnar layout opens up compression techniques that aren’t available in MVT’s row-oriented format: delta encoding for coordinates, run-length encoding for repeated values, and dictionary encoding for string columns with low cardinality.
The practical result is smaller tiles, especially for polygon-heavy datasets:
| Dataset | MVT | MLT | Savings |
|---|---|---|---|
| NC counties (z0-14) | 78 KB | 65 KB | 17% |
| US block groups (z4-12) | 310 MB | 259 MB | 16% |
| Education dots (z0-12) | 238 MB | 237 MB | ~0% |
Point-only datasets see minimal savings because coordinate data dominates and compresses similarly in both formats.
Which format should I use?
Use MLT (the default) when you’re working in R with mapgl or viewing in a browser with MapLibre GL JS 5.17+. MLT produces smaller files for polygon and line data, and mapgl supports it natively.
Use MVT when you need the widest viewer compatibility - particularly for Python-facing examples, older MapLibre versions, or when sharing tilesets with users on tools you don’t control. You can switch formats with a single argument:
Ecosystem status
MLT is still new (January 2026), and library support is growing:
- Encoding: freestiler (Rust, R + Python), mlt-core reference encoder (Java/C++)
- Decoding: MapLibre GL JS (experimental), mlt-core decoder
- Viewing: MapLibre GL JS with MLT plugin, mapgl in R
As MLT support expands across mapping libraries, the default will become directly viewable in more places without needing to switch to MVT.