runs on your GPU · nothing uploaded
Easy advice at a thousand rows, hard at a hundred million. QPlot draws every row and keeps pan, zoom and recolor at full frame rate. Try a dataset on the right, or drop yours.
Your file stays on your machine. No account needed.
Real rows from the demo file. This preview draws a 12,000-row sample on a 2D canvas; QPlot draws all of them on the GPU.
Each opens as a live document. Pan, zoom, recolor, facet, hover a point. Nothing is sampled.
Read only what the plot needs, describe the plot with a grammar instead of a chart menu, and compute the statistics on the GPU so they keep up with the view.
Parquet, Arrow and CSV are read column by column, in ranges, from disk or a URL. Columns you never bind are never loaded, so a 100M-row file opens in seconds.
Bind columns to x, y, color, size and facet; choose points, lines, density or images as the layer. Swap one binding and the same rows re-plot, as the preview above does.
Histograms, 2D density, smoothing and correlation are computed on the GPU for whatever is in view, and recomputed as you pan and zoom.
Add an expression, bin a value, reorder a categorical, or filter one plot without touching the others. The results are columns like any other.
Save views as slides, build a deck, share a link. Teammates open the live document and keep exploring. The data stays where it already lives.
WebGPU in the browser or the desktop app on Metal. Local files never leave the device; URL sources stream straight to the GPU. There is no upload step.
[1–12] The headline is a quotation. Wilkinson & the APA Task Force, 1999: “before you compute any statistics, look at your data.” Eleven more say the same thing, 1925 to 2019.
This is taking longer than usual — the app is about 5 MB. It will open as soon as it arrives.
Loading…