Python GUI benchmarks — Pyzed vs tkinter, Qt, Kivy, Flet…

These numbers were measured, not estimated: every chart on this page is drawn from the JSON file the benchmark script wrote, and you can run the same script on your machine. They come from one machine without a GPU, so read them as a comparison between toolkits under the same conditions, not as absolute numbers for your computer.

Software rendering

The machine has no GPU. Windows were shown on a virtual X server (Xvfb) and every toolkit that uses OpenGL or Vulkan (Pyzed, Kivy, Dear PyGui, Flet) rendered on the CPU through Mesa’s llvmpipe and lavapipe drivers. That is the worst case for GPU toolkits and the normal case for tkinter, Qt Widgets and GTK, which draw on the CPU anyway. On a computer with a GPU, the drawing part of the GPU toolkits’ numbers moves off the CPU.

Key findings

• Startup is fast everywhere except Flet and, with many widgets, Kivy and Toga. Pyzed shows a 10-row window in about 0.16 s from launch and a 1,000-row window in about 0.24 s: building and laying out 2,000 more widgets costs it some 80 ms. tkinter, Dear PyGui and Qt start sooner; Kivy, Toga and Flet need close to a second for 1,000 rows.
• Updating 1,000 rows takes Pyzed about 18 ms from state change to screen, even though it rebuilds the whole view from Python each time. That is a little faster than Qt Widgets, about 3× faster than Kivy and 9× faster than Flet; tkinter and Dear PyGui are faster still. Toga (GTK) needed about 37 s per update in this test, because every label change triggers a new layout.
• Memory depends on the renderer. Through the software Vulkan driver used here, Pyzed’s GPU renderer takes about 151 MB; the same app on Pyzed’s CPU renderer (tiny-skia, set with ICED_BACKEND=tiny-skia) takes about 48 MB, between tkinter (25 MB) and Qt (72 MB). The other GPU toolkits pay a similar driver cost here (Dear PyGui 123 MB, Kivy 146 MB, Flet 268 MB).
• Install size: about 30 MB for Pyzed, against 238 MB for Qt’s essentials, 83 MB for Kivy and 51 MB for Flet with its client. tkinter, Dear PyGui, Toga (which uses the system’s GTK) and CustomTkinter (which uses tkinter) are smaller.
• Code: the to-do app takes 31 lines in Pyzed, half of PySide6 (62) and well under Kivy (81). Flet’s is the shortest (22).

Startup

Import is the time to import the toolkit in a fresh Python process. First frame is the time from launching the app’s process until its window is on the screen with its content: the script watches the screen for a colored bar at the bottom of the window.

Import the toolkitms · lower is better

1. tkinterTk 9.0.35.3 ms
2. Dear PyGui2.3.117 ms
3. Toga0.5.729 ms
4. Pyzed1.2.035 ms
5. Flet1.0.440 ms
6. CustomTkinter6.0.040 ms
7. PySide6 (Qt)6.12.068 ms
8. Kivy2.3.199 ms

Launch to first frame: window with 10 rowsms · lower is better

1. tkinterTk 9.0.396 ms
2. Pyzed (CPU renderer)1.2.0 · tiny-skia118 ms
3. Dear PyGui2.3.1133 ms
4. PySide6 (Qt)6.12.0135 ms
5. CustomTkinter6.0.0162 ms
6. Pyzed1.2.0165 ms
7. Toga0.5.7221 ms
8. Kivy2.3.1258 ms
9. Flet1.0.4579 ms

Launch to first frame: window with 1,000 rowsms · lower is better

1. Dear PyGui2.3.1152 ms
2. tkinterTk 9.0.3157 ms
3. Pyzed (CPU renderer)1.2.0 · tiny-skia175 ms
4. PySide6 (Qt)6.12.0217 ms
5. Pyzed1.2.0243 ms
6. Toga0.5.7887 ms
7. Flet1.0.4994 ms
8. Kivy2.3.11,101 ms

One real widget per row (a label and a button), no virtualized lists. CustomTkinter is missing: its 1,000-row app crashed (see the caveats).

Memory

Resident memory (RSS) of the app’s processes one second after the first frame. For Flet it includes the Flutter client process that draws the window.

Memory: window with 10 rowsMB · lower is better

1. tkinterTk 9.0.325 MB
2. CustomTkinter6.0.031 MB
3. Pyzed (CPU renderer)1.2.0 · tiny-skia48 MB
4. Toga0.5.758 MB
5. PySide6 (Qt)6.12.072 MB
6. Dear PyGui2.3.1123 MB
7. Kivy2.3.1146 MB
8. Pyzed1.2.0151 MB
9. Flet1.0.4268 MB

Memory: window with 1,000 rowsMB · lower is better

1. tkinterTk 9.0.334 MB
2. Pyzed (CPU renderer)1.2.0 · tiny-skia57 MB
3. PySide6 (Qt)6.12.083 MB
4. Toga0.5.785 MB
5. Dear PyGui2.3.1125 MB
6. Pyzed1.2.0161 MB
7. Kivy2.3.1282 MB
8. Flet1.0.4395 MB

Updating 1,000 rows

Every 250 ms the app changes the text of all 1,000 labels and the color of the bottom bar in the same update. The latency is the time from the app changing its state to the new color being on the screen; 10 updates per launch, 7 launches.

Update 1,000 labels: state change to screenms · lower is better

1. Dear PyGui2.3.16.6 ms
2. tkinterTk 9.0.37.7 ms
3. Pyzed (CPU renderer)1.2.0 · tiny-skia15 ms
4. Pyzed1.2.018 ms
5. PySide6 (Qt)6.12.024 ms
6. Kivy2.3.159 ms
7. Flet1.0.4166 ms
8. Toga0.5.736,619 ms

The axis is cut at 250 ms: Toga's bar is broken and its real value is shown on the right.

Update 10 labels: state change to screenms · lower is better

1. tkinterTk 9.0.31.9 ms
2. Pyzed (CPU renderer)1.2.0 · tiny-skia2.7 ms
3. CustomTkinter6.0.02.8 ms
4. PySide6 (Qt)6.12.02.8 ms
5. Dear PyGui2.3.13.5 ms
6. Pyzed1.2.03.8 ms
7. Kivy2.3.14.1 ms
8. Toga0.5.79.1 ms
9. Flet1.0.425 ms

Install size

How much a fresh virtual environment grows when the toolkit is installed with its dependencies. tkinter counts the Tcl/Tk files that come with Python. Flet includes the desktop client it downloads on first run. Toga does not include GTK, which comes from the operating system. NiceGUI is shown for reference only: it draws in a web browser, so it was not part of the window benchmarks.

Installed size, with dependenciesMB · lower is better

1. CustomTkinter6.0.01.4 MB
2. Toga0.5.72.9 MB
3. tkinterTk 9.0.39.3 MB
4. Dear PyGui2.3.110.0 MB
5. Flet1.0.451 MB
6. NiceGUI3.18.077 MB
7. Kivy2.3.183 MB
8. PySide6 (Qt)6.12.0238 MB

Lines of code

The same counter and to-do apps written in each toolkit (source), counting lines that are not blank, comments or docstrings. See Developer experience for the charts and what the lines do.

Raw numbers

Median in bold; below it the minimum–maximum and the number of samples.

Library	Import (ms)	First frame, 10 rows (ms)	First frame, 1,000 rows (ms)	RSS, 10 rows (MB)	RSS, 1,000 rows (MB)
Pyzed 1.2.0	35 34–70 · n=10	165 154–691 · n=7	243 188–363 · n=7	151 150–158 · n=7	161 160–161 · n=7
tkinter Tk 9.0.3	5.3 5.1–5.7 · n=10	96 85–100 · n=7	157 152–186 · n=7	25 25–25 · n=7	34 33–34 · n=7
CustomTkinter 6.0.0	40 40–42 · n=10	162 152–181 · n=7	–	31 31–31 · n=7	–
PySide6 (Qt) 6.12.0	68 65–76 · n=10	135 127–151 · n=7	217 204–247 · n=7	72 72–72 · n=7	83 83–83 · n=7
Kivy 2.3.1	99 96–133 · n=10	258 249–276 · n=7	1,101 1,022–1,153 · n=7	146 146–147 · n=7	282 281–282 · n=7
Dear PyGui 2.3.1	17 17–19 · n=10	133 123–139 · n=7	152 145–178 · n=7	123 123–123 · n=7	125 125–125 · n=7
Flet 1.0.4	40 38–42 · n=10	579 543–589 · n=7	994 923–1,007 · n=7	268 265–270 · n=7	395 392–403 · n=7
Toga 0.5.7	29 28–30 · n=10	221 216–259 · n=7	887 837–955 · n=7	58 58–58 · n=7	85 85–85 · n=7
Pyzed (CPU renderer) 1.2.0 · tiny-skia	–	118 94–163 · n=7	175 118–201 · n=7	48 47–48 · n=7	57 57–58 · n=7

Library	Update, 10 rows (ms)	Update, 1,000 rows (ms)	Install size (MB)	Counter (lines)	To-do (lines)
Pyzed 1.2.0	3.8 2.4–21 · n=70	18 12–46 · n=70	–	19	31
tkinter Tk 9.0.3	1.9 1.1–5.9 · n=70	7.7 6.6–12 · n=70	9.3	15	39
CustomTkinter 6.0.0	2.8 1.7–5.4 · n=70	–	1.4	14	32
PySide6 (Qt) 6.12.0	2.8 1.8–9.2 · n=70	24 21–38 · n=70	238	28	62
Kivy 2.3.1	4.1 2.9–12 · n=70	59 46–259 · n=70	83	31	81
Dear PyGui 2.3.1	3.5 2.6–6.3 · n=70	6.6 4.0–9.2 · n=70	10.0	19	29
Flet 1.0.4	25 14–40 · n=70	166 145–194 · n=70	51	17	22
Toga 0.5.7	9.1 8.4–14 · n=70	36,619 36,055–37,464 · n=70	2.9	19	35
Pyzed (CPU renderer) 1.2.0 · tiny-skia	2.7 1.5–12 · n=70	15 13–50 · n=70	–	–	–
NiceGUI 3.18.0	–	–	77	–	–

The complete results, with every sample, are in benchmarks/results/.

Machine and versions

• Date: 2026-10-19
• OS: Linux-6.18.44-fc-v139-x86_64-with-glibc2.39
• CPU: Intel(R) Xeon(R) Processor, 4 cores; 15.7 GB of RAM (a cloud container)
• Display: Xvfb (X virtual framebuffer, no GPU); OpenGL/Vulkan through Mesa llvmpipe/lavapipe; Mesa 25.2.8-0ubuntu0.24.04.4
• Python 3.13.14 (python-build-standalone, via uv); Tk 9.0.3

• Toolkits:

  Pyzed 1.2.0, tkinter Tk 9.0.3, CustomTkinter 6.0.0, PySide6 (Qt) 6.12.0, Kivy 2.3.1, Dear PyGui 2.3.1, Flet 1.0.4, Toga 0.5.7

Method

• Same app everywhere. Each toolkit’s bench.py builds a 640×480 window with a scrollable list of N rows (a label and an “Open” button each) and a 40 px bar at the bottom painted in a sentinel color. Lists are built the plain way each toolkit documents, with one widget per row: no RecycleView in Kivy, no lazy ListView in Flet, no virtual_list in Pyzed.
• First frame is measured from outside: the harness starts the process and grabs the screen about every millisecond (with mss) until 8,000 pixels of the sentinel color are visible.
• Memory is the RSS of the process and its children (psutil), one second after the first frame.
• Updates are timed between the app’s own timestamp (time.monotonic(), printed right before it changes its state) and the moment the harness sees the bar’s new color.
• Repetitions: 7 launches per toolkit and size, 10 updates per launch, 10 imports; the charts show the median and a thin line from the minimum to the maximum.
• Pyzed (CPU renderer) is the same Pyzed app started with ICED_BACKEND=tiny-skia, which makes Iced use its CPU renderer instead of wgpu: what Pyzed falls back to on machines without a working GPU driver.
• Install size: a fresh uv venv, uv pip install --link-mode=copy <package>, size difference.

Reproduce it

uv

macOS and Linux

git clone https://github.com/albakore/pyzed
cd pyzed
uv venv --python 3.13 .venv
source .venv/bin/activate
uv pip install pyzed-ui pyside6 kivy dearpygui "flet[desktop]" toga customtkinter psutil mss numpy
python benchmarks/run.py

Windows (PowerShell)

git clone https://github.com/albakore/pyzed
cd pyzed
uv venv --python 3.13 .venv
.venv\Scripts\Activate.ps1
uv pip install pyzed-ui pyside6 kivy dearpygui "flet[desktop]" toga customtkinter psutil mss numpy
python benchmarks/run.py

pip

macOS and Linux

git clone https://github.com/albakore/pyzed
cd pyzed
python3 -m venv .venv
source .venv/bin/activate
pip install pyzed-ui pyside6 kivy dearpygui "flet[desktop]" toga customtkinter psutil mss numpy
python benchmarks/run.py

Windows (PowerShell)

git clone https://github.com/albakore/pyzed
cd pyzed
py -m venv .venv
.venv\Scripts\Activate.ps1
pip install pyzed-ui pyside6 kivy dearpygui "flet[desktop]" toga customtkinter psutil mss numpy
python benchmarks/run.py

Without a display it starts Xvfb by itself; on a desktop, keep the benchmark windows uncovered while it runs. See benchmarks/README.md for the options and the system packages some toolkits need. Results on other machines are very welcome in an issue.

Caveats

• One machine, no GPU (see above). Results on Windows and macOS, and with a real GPU, will differ.
• The harness shares the CPU with the app: grabbing the screen costs about one core while it polls.
• Tk draws widget by widget, so for tkinter the bar may appear a moment before every row has been drawn; the GPU toolkits present whole frames.
• Dear PyGui redraws continuously, so its update latency includes waiting for its next frame; it also uses the CPU while idle, which is not measured here.
• CustomTkinter has no 1,000-row results: its app died under Xvfb with an X error (BadAlloc or BadValue from CreatePixmap) once its scrollable frame held more than about 100 rows.
• Not measured: wxPython has no Linux wheels on PyPI and building it from source was not practical here; NiceGUI runs in a web browser, so a desktop window timing would measure the browser.
• Lines of code depend on style; the apps follow each toolkit’s documented idioms, and improvements are welcome.