Performance

Every example in the plot-type reference gallery serialized to static SVG and to self-contained interactive HTML, with output sizes. (The real applications gallery is not timed here: its figures are variations on the same shapes, and timing another hundred of them would treble the runtime without adding a row that says anything new.) Interactive HTML embeds the same SVG plus the per-axes data the toolbar needs for zoom / point-picking (the picked values, and mesh z grids), so it is larger and slower than SVG – most for mesh-heavy figures.

Best of 5 runs, one machine. Regenerate with python benchmarks/example_timings.py.

Against other libraries

The same four figures built and serialized to SVG by each library, using its own idiomatic API and no global state on any side. matplotlib goes through the object-oriented FigureCanvasSVG rather than pyplot; xy renders headlessly through Chart.to_svg(). xy facets by a data column rather than by an arbitrary grid, so its 8x8 case is 64 groups of one long-form table – the idiomatic equivalent, not a handicap.

These measure time to produce a static file, which is the axis plotpress optimizes. It is not the axis xy optimizes: its Rust core decimates by screen resolution for interactive exploration of large data, and a single static render does not exercise that.

Scenario	plotpress (time / size)	matplotlib (time / size)	xy (time / size)
line_100k_points	8.8 ms / 17 KiB	41.1 ms / 24 KiB	2.5 ms / 31 KiB
scatter_5k_points	14.9 ms / 137 KiB	118.1 ms / 533 KiB	22.2 ms / 360 KiB
pcolormesh_300x300	9.2 ms / 45 KiB	5661.0 ms / 15.4 MiB	15.6 ms / 95 KiB
many_axes_8x8_grid	36.1 ms / 336 KiB	1317.6 ms / 331 KiB	58.9 ms / 361 KiB

Per-example output

Example	Axes	SVG	SVG size	HTML	HTML size
pairwise/plot_01_line	1	1.6 ms	10 KiB	3.0 ms	85 KiB
pairwise/plot_02_scatter	1	3.1 ms	44 KiB	3.9 ms	119 KiB
pairwise/plot_03_bar	1	0.5 ms	4 KiB	0.9 ms	75 KiB
pairwise/plot_04_barh	1	0.4 ms	4 KiB	0.8 ms	75 KiB
pairwise/plot_05_stem	1	0.3 ms	5 KiB	0.6 ms	76 KiB
pairwise/plot_06_step	1	0.3 ms	3 KiB	0.6 ms	73 KiB
pairwise/plot_07_fill_between	2	4.0 ms	21 KiB	5.6 ms	104 KiB
pairwise/plot_08_stackplot	1	0.5 ms	4 KiB	0.9 ms	75 KiB
pairwise/plot_09_loglog	3	2.2 ms	14 KiB	3.4 ms	88 KiB
pairwise/plot_10_reference_lines	1	1.8 ms	16 KiB	3.0 ms	94 KiB
pairwise/plot_11_broken_barh	1	0.4 ms	3 KiB	0.9 ms	74 KiB
pairwise/plot_12_stairs	1	0.3 ms	3 KiB	0.6 ms	74 KiB
pairwise/plot_13_axline	1	0.4 ms	4 KiB	0.9 ms	75 KiB
pairwise/plot_14_hlines_vlines	2	1.6 ms	14 KiB	2.8 ms	92 KiB
pairwise/plot_15_linestyles	2	3.8 ms	29 KiB	6.5 ms	116 KiB
distributions/plot_01_hist	1	1.0 ms	8 KiB	1.4 ms	80 KiB
distributions/plot_02_boxplot	1	0.3 ms	4 KiB	0.6 ms	75 KiB
distributions/plot_03_errorbar	1	0.4 ms	5 KiB	0.6 ms	75 KiB
distributions/plot_04_violin	1	1.5 ms	10 KiB	2.7 ms	85 KiB
distributions/plot_05_eventplot	1	1.2 ms	16 KiB	1.9 ms	88 KiB
distributions/plot_06_hist2d	1	0.7 ms	6 KiB	1.7 ms	82 KiB
distributions/plot_07_pie	1	0.1 ms	1 KiB	0.3 ms	72 KiB
distributions/plot_08_hexbin	1	13.8 ms	63 KiB	22.4 ms	151 KiB
gridded_data/plot_01_imshow	1	1.7 ms	17 KiB	6.3 ms	164 KiB
gridded_data/plot_02_pcolormesh	1	17.7 ms	50 KiB	40.3 ms	544 KiB
gridded_data/plot_03_contour	1	32.4 ms	87 KiB	37.0 ms	235 KiB
gridded_data/plot_04_quiver	1	2.1 ms	27 KiB	3.1 ms	104 KiB
gridded_data/plot_05_contourf	1	18.2 ms	12 KiB	82.9 ms	1.3 MiB
gridded_data/plot_06_curvilinear_mesh	1	290.3 ms	41 KiB	283.9 ms	149 KiB
gridded_data/plot_07_gouraud	2	184.7 ms	289 KiB	184.0 ms	372 KiB
gridded_data/plot_08_lognorm	2	7.0 ms	29 KiB	29.5 ms	521 KiB
gridded_data/plot_09_matshow	1	0.6 ms	7 KiB	1.0 ms	78 KiB
gridded_data/plot_10_spy	1	0.4 ms	3 KiB	1.0 ms	77 KiB
gridded_data/plot_11_colormap_reference	5	4.0 ms	42 KiB	11.6 ms	180 KiB
multi_axes/plot_01_subplots	4	2.5 ms	24 KiB	4.7 ms	111 KiB
multi_axes/plot_02_annotations	1	0.6 ms	6 KiB	1.2 ms	78 KiB
multi_axes/plot_03_twin_axes	4	3.1 ms	24 KiB	5.1 ms	110 KiB
multi_axes/plot_04_shared_colorbar	6	3.7 ms	22 KiB	12.2 ms	213 KiB
animation/plot_01_plot_frames	1	0.7 ms	6 KiB	4.8 ms	110 KiB
animation/plot_02_pcolormesh_frames	4	3.8 ms	23 KiB	246.8 ms	3.7 MiB
scale/plot_01_many_axes	500	395.9 ms	2.7 MiB	836.3 ms	7.5 MiB
scale/plot_02_million_point_line	1	156.8 ms	47 KiB	184.9 ms	117 KiB
scale/plot_03_multimillion_mesh	1	320.6 ms	775 KiB	419.8 ms	2.1 MiB
scale/plot_04_thousand_series	1	887.8 ms	6.9 MiB	1362.2 ms	12.1 MiB
scale/plot_05_axes_grid_lines	900	20173.4 ms	2.4 MiB	58817.4 ms	3.9 MiB
scale/plot_06_shared_colorbar_lognorm	144	81.5 ms	514 KiB	236.8 ms	2.5 MiB
scale/plot_07_matplotlib_comparison	1	0.9 ms	5 KiB	1.3 ms	76 KiB
scale/plot_08_vector_over_raster	1	266.3 ms	600 KiB	437.4 ms	3.2 MiB
scale/plot_09_output_scaling	1	1.6 ms	7 KiB	2.3 ms	79 KiB
scale/plot_10_many_groups_with_colorbars	1	0.3 ms	3 KiB	0.6 ms	74 KiB

The scale/plot_01_many_axes row is the deliberate stress case: 500 independent pcolormesh axes on one figure. Its interactive HTML is dominated by the 500 embedded mesh z grids; lower fig.to_html(pick_precision=...) (or fig.save(..., pick_precision=...)) to trade readout precision for a smaller file, or cap the total embedded amount directly with pick_max_mesh_cells / pick_max_points – a mesh over the cap is block-averaged down to it rather than dropped, so a click still answers with a real, if coarser, value.

Binary vs. JSON pick data

fig.to_html()/fig.save(...html) embed long point-pick arrays (mesh z grids, animated line frames) as base64 float32 bytes by default (binary_pick_data=True) rather than JSON number text. Below is every example above, both ways: most are small enough that neither the array-length threshold nor the file size difference matters: the JSON version is well under a point-pick array’s own threshold to begin with. It shows up once a mesh or a long series does most of the work, which is exactly the scale/ rows.

Example	Binary	Binary size	JSON	JSON size	Size	Time
pairwise/plot_01_line	3.0 ms	85 KiB	2.7 ms	88 KiB	1.04x	0.88x
pairwise/plot_02_scatter	3.9 ms	119 KiB	3.5 ms	123 KiB	1.03x	0.89x
pairwise/plot_03_bar	0.9 ms	75 KiB	0.8 ms	75 KiB	0.99x	0.99x
pairwise/plot_04_barh	0.8 ms	75 KiB	0.7 ms	75 KiB	0.99x	0.81x
pairwise/plot_05_stem	0.6 ms	76 KiB	0.5 ms	76 KiB	0.99x	0.95x
pairwise/plot_06_step	0.6 ms	73 KiB	0.5 ms	73 KiB	0.99x	0.92x
pairwise/plot_07_fill_between	5.6 ms	104 KiB	5.1 ms	112 KiB	1.08x	0.90x
pairwise/plot_08_stackplot	0.9 ms	75 KiB	0.9 ms	75 KiB	0.99x	0.92x
pairwise/plot_09_loglog	3.4 ms	88 KiB	2.9 ms	91 KiB	1.03x	0.85x
pairwise/plot_10_reference_lines	3.0 ms	94 KiB	2.7 ms	99 KiB	1.05x	0.93x
pairwise/plot_11_broken_barh	0.9 ms	74 KiB	0.8 ms	73 KiB	0.99x	0.93x
pairwise/plot_12_stairs	0.6 ms	74 KiB	0.5 ms	73 KiB	0.99x	0.90x
pairwise/plot_13_axline	0.9 ms	75 KiB	0.7 ms	75 KiB	1.00x	0.78x
pairwise/plot_14_hlines_vlines	2.8 ms	92 KiB	2.7 ms	97 KiB	1.06x	0.95x
pairwise/plot_15_linestyles	6.5 ms	116 KiB	6.1 ms	129 KiB	1.11x	0.93x
distributions/plot_01_hist	1.4 ms	80 KiB	1.3 ms	80 KiB	1.00x	0.93x
distributions/plot_02_boxplot	0.6 ms	75 KiB	0.5 ms	74 KiB	0.99x	0.85x
distributions/plot_03_errorbar	0.6 ms	75 KiB	0.6 ms	75 KiB	0.99x	0.90x
distributions/plot_04_violin	2.7 ms	85 KiB	2.1 ms	88 KiB	1.03x	0.79x
distributions/plot_05_eventplot	1.9 ms	88 KiB	1.7 ms	89 KiB	1.01x	0.90x
distributions/plot_06_hist2d	1.7 ms	82 KiB	1.4 ms	85 KiB	1.05x	0.82x
distributions/plot_07_pie	0.3 ms	72 KiB	0.2 ms	72 KiB	0.99x	0.75x
distributions/plot_08_hexbin	22.4 ms	151 KiB	22.5 ms	166 KiB	1.11x	1.00x
gridded_data/plot_01_imshow	6.3 ms	164 KiB	8.8 ms	234 KiB	1.43x	1.40x
gridded_data/plot_02_pcolormesh	40.3 ms	544 KiB	53.6 ms	871 KiB	1.60x	1.33x
gridded_data/plot_03_contour	37.0 ms	235 KiB	39.5 ms	308 KiB	1.31x	1.07x
gridded_data/plot_04_quiver	3.1 ms	104 KiB	2.8 ms	108 KiB	1.04x	0.92x
gridded_data/plot_05_contourf	82.9 ms	1.3 MiB	126.0 ms	2.4 MiB	1.88x	1.52x
gridded_data/plot_06_curvilinear_mesh	283.9 ms	149 KiB	281.6 ms	184 KiB	1.23x	0.99x
gridded_data/plot_07_gouraud	184.0 ms	372 KiB	184.9 ms	382 KiB	1.03x	1.00x
gridded_data/plot_08_lognorm	29.5 ms	521 KiB	45.4 ms	927 KiB	1.78x	1.54x
gridded_data/plot_09_matshow	1.0 ms	78 KiB	0.9 ms	78 KiB	1.00x	0.87x
gridded_data/plot_10_spy	1.0 ms	77 KiB	0.8 ms	78 KiB	1.02x	0.81x
gridded_data/plot_11_colormap_reference	11.6 ms	180 KiB	10.9 ms	232 KiB	1.28x	0.95x
multi_axes/plot_01_subplots	4.7 ms	111 KiB	4.5 ms	125 KiB	1.12x	0.97x
multi_axes/plot_02_annotations	1.2 ms	78 KiB	1.2 ms	80 KiB	1.02x	0.95x
multi_axes/plot_03_twin_axes	5.1 ms	110 KiB	4.9 ms	121 KiB	1.10x	0.96x
multi_axes/plot_04_shared_colorbar	12.2 ms	213 KiB	15.2 ms	306 KiB	1.44x	1.25x
animation/plot_01_plot_frames	4.8 ms	110 KiB	4.0 ms	140 KiB	1.27x	0.82x
animation/plot_02_pcolormesh_frames	246.8 ms	3.7 MiB	352.0 ms	6.1 MiB	1.64x	1.43x
scale/plot_01_many_axes	836.3 ms	7.5 MiB	919.7 ms	11.7 MiB	1.55x	1.10x
scale/plot_02_million_point_line	184.9 ms	117 KiB	184.5 ms	117 KiB	1.00x	1.00x
scale/plot_03_multimillion_mesh	419.8 ms	2.1 MiB	467.5 ms	3.3 MiB	1.58x	1.11x
scale/plot_04_thousand_series	1362.2 ms	12.1 MiB	1460.5 ms	17.3 MiB	1.42x	1.07x
scale/plot_05_axes_grid_lines	58817.4 ms	3.9 MiB	58642.3 ms	5.1 MiB	1.29x	1.00x
scale/plot_06_shared_colorbar_lognorm	236.8 ms	2.5 MiB	285.1 ms	4.1 MiB	1.69x	1.20x
scale/plot_07_matplotlib_comparison	1.3 ms	76 KiB	1.3 ms	76 KiB	0.99x	0.95x
scale/plot_08_vector_over_raster	437.4 ms	3.2 MiB	531.2 ms	5.6 MiB	1.75x	1.21x
scale/plot_09_output_scaling	2.3 ms	79 KiB	2.2 ms	78 KiB	0.99x	0.97x
scale/plot_10_many_groups_with_colorbars	0.6 ms	74 KiB	0.5 ms	74 KiB	0.99x	0.90x

“Size”/”Time” are JSON relative to binary – 2.0x under Size means the JSON payload is twice the binary one’s size; under Time means it took twice as long to build. A ratio near 1.0x on a small figure means the encoder found nothing worth switching: every array in it was already under the threshold where a base64 wrapper costs more than it saves.