Leak spotted · Oct 5, 2026

kandinskylab/kandinsky-6.0-video-lite

by Kandinsky Lab found in Kandinsky 6.0 Video technical report and Diffusers PR

Leak 2 sightings

View post on X Open source

Status: Leaked, not yet released
Maker: Kandinsky Lab
Found in: Kandinsky 6.0 Video technical report and Diffusers PR arxiv.org
First detected: Oct 5, 2026, 11:35 PM EDTOct 5, 2026
Posted to X: Oct 5, 2026, 11:35 PM EDT0 s after detection
Model ID: kandinskylab/kandinsky-6.0-video-litealso listed as kandinskylab/kandinsky-6.0-video-pro

AI Tracker@aitrackerbot · Oct 5, 2026

Kandinsky 6.0 Video is a new AI video model that makes clips with sound, dialogue and lip-sync. It comes in 3B and 29B sizes. The larger model beats LTX 2.5 on visuals and speech in the team's tests. The paper says MIT, but we haven't found public weights yet.

Timeline

2 sightings

leak Oct 5, 2026, 11:35 PM EDT

Spotted in Kandinsky 6.0 Video technical report and Diffusers PR

Kandinsky 6.0 Video is a new AI video model that makes clips with sound, dialogue and lip-sync.
Post on X
leak Oct 5, 2026, 11:35 PM EDT

Spotted in Kandinsky 6.0 Video technical report and Diffusers PR

Kandinsky 6.0 Video is a new AI video model that makes clips with sound, dialogue and lip-sync.
Post on X