Kandinsky 6.0 Pro and Kandinsky 6.0 Lite
Kandinsky is Kandinsky Lab's open model line. Kandinsky 6.0, released on 6 October 2026 under the MIT licence, is two video models that make five-second clips with synchronized audio: Pro at 29 billion parameters and Lite at 3 billion, each with a distilled variant, plus a super-resolution model that lifts the 480p base clip to 1080p. fal serves both at 480p with optional upscaling, billed per clip by step count. Both pages run the same nine clips as every other video model here.
Kandinsky Lab · facts checked
Which one to pick
Kandinsky 6.0 Pro
Open-weight clips under MIT that follow the prompt, including a faithful product photo, if you can wait 14 to 26 minutes a clip on fal.
Kandinsky 6.0 Lite
$0.16 per five-second 480p clip, text-to-video via fal
The smaller model, to run yourself under MIT; on fal, check every clip, since three of ours showed something else.
Side by side, measured
| Model | Price | Seconds to make one clip | 9 clips cost | Length |
|---|---|---|---|---|
| Kandinsky 6.0 Pro | 830 to 1553 | $12.20 | Five seconds; there is no duration setting | |
| Kandinsky 6.0 Lite | $0.16 per five-second 480p clip, text-to-video via fal | 193 to 777 | $1.45 | Five seconds; there is no duration setting |
Seconds and cost are from our own runs of the same 9 prompts through each model; open a model for every clip and the request behind it.
Every version
| Version | Released | What changed |
|---|---|---|
| Kandinsky 6.0 Pro fal-ai/kandinsky6-pro | The 29B model: five-second 480p clips with 44 kHz audio, MIT licensed. $1.35 a clip on fal at the default 50 steps. | |
| Kandinsky 6.0 Lite fal-ai/kandinsky6-lite | The 3B model from the same release, tuned for 10 steps: $0.16 a clip on fal. |
Frequently asked questions
Is Kandinsky 6.0 open source?
Yes. Kandinsky Lab published code and weights for Pro and Lite, their distilled variants and the super-resolution model under the MIT licence on 6 October 2026, on GitHub and Hugging Face.
Who makes Kandinsky?
Kandinsky Lab, the team behind the Kandinsky image and video models; the GitHub organization is kandinskylab.