whatmodel.sh / video / Kandinsky 5.0 T2V Lite

Kandinsky 5.0 T2V Lite

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Kandinsky 5.0 T2V Lite is an open-weights video model from Sber, released September 2025. Its published size is 2B parameters. Licence: MIT.

Model information sources

Release date ↗ · Parameter count ↗

2B as the maker states for the T2V Lite model (the generative DiT line; text encoders Qwen2.5-VL/CLIP and HunyuanVideo VAE are separate components not counted explicitly).

47.0±2.4#58 of 93 video modelsfrom 1 benchmark result across 1 source* thinly tested, treat the rank as provisional

$ how it compares

The video models ranked either side of Kandinsky 5.0 T2V Lite, out of 139 with a score, then more from Sber.

#modelscorewhereRAMreleased
80Wan 2.1 14B Alibaba47±2api $0.24/sec +local~20GB2025-02
81Wan 2.2 TI2V 5B Alibaba47±2api price unknown +local~6.5GB2025-07
82Kandinsky 5.0 T2V Lite Sber47±2local~29GB2025-09
83LTX-2 19B Distilled Lightricks47±8api $0.020/sec +local~41GB2026-01
84Cosmos-Predict2.5-14B NVIDIA47±3comparison only2025-12
more from Sber
57Kandinsky 5.0 Pro Sber52±2api $0.040/sec +local~76GB2025-11
64Kandinsky 6.0 Video Lite Sber51±7local~34GB2026-10
77Kandinsky 6.0 Video Pro Sber48±7local~109GB2026-10

see all video models · find the ones that fit your machine

$ what it's good at

text to video▆▄▃▄▆▅█▃▂▂▂▄ #37 of 68 * · 98% of its score · 1 test

Beats Wan 2.2 TI2V 5B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
LMArena video1112.6 ✎▂▂▃▆█▆·▆▅▁▂▇ beats 4/41avg 1147▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 98%typical

$ where to run it

Other ways to run it

  1. Run it as published · needs ~29.1 GB, NVIDIA only · from huggingface.co/kandinskylab
    uv run --with torch --with diffusers --with transformers --with accelerate --with imageio-ffmpeg python -c "import torch; from diffusers import Kandinsky5T2VPipeline; from diffusers.utils import export_to_video; p=Kandinsky5T2VPipeline.from_pretrained('kandinskylab/Kandinsky-5.0-T2V-Lite-sft-5s-Diffusers', torch_dtype=torch.bfloat16).to('cuda'); export_to_video(p(prompt='A cat and a dog baking a cake together in a kitchen.', negative_prompt='Static, 2D cartoon, cartoon, 2d animation, paintings, images, worst quality, low quality, ugly, deformed, walking backwards', height=512, width=768, num_frames=121, num_inference_steps=50, guidance_scale=5.0).frames[0], 'output.mp4', fps=24, quality=9)"

Weights and documentation: github.com/kandinskylab ↗

$ how the score was worked out · LMArena leaderboard dataset 98%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 98%CC BY 4.0
prior · pull toward the middle▇ 2%
weighted average= 47.0

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json