Skip to main content

AlibabaVideo model

Lumos-1 3B SFT 384p

Clip lengthn/a

n/alongest single cliphigher is better

Summary

Lumos-1 3B SFT 384p is a video model from Alibaba. One shared autoregressive checkpoint supports text-to-video and image-to-video. These capabilities are not counted as separate model rows.

What Lumos-1 3B SFT 384p can do

“n/a” means the provider does not state it or we have not sourced it yet. That is not a “no”.

Text to videoyesSource
Image to videoyesSource
Referencesn/a
AudionoSource
Runs locallyyesSource
Extendn/a
Start and end framen/a
Max. resolution384pSource

Head-to-head comparisons

Show all comparisons (94)

All specifications

Architecture

Parameters
3B
Source: Parameters
Current checkpoint
The latest released 3B weights are `3B/v2-stage2-joint-384p-sft`; the vendor lists them separately from stage-1 and earlier stage-2 checkpoints.
Source: Current checkpoint

Input and output

Maximum resolution
384p
Source: Maximum resolution
Native audio
No
Source: Native audio

Licence and access

Weights
Open weights
Source: Weights
Access
Local
Source: Access
Release
Alibaba DAMO dates the SFT checkpoint to March 2026.
Source: Release

Further details

Input modes
One shared autoregressive checkpoint supports text-to-video and image-to-video. These capabilities are not counted as separate model rows.
Source: Input modes
Weights and use
Code and checkpoints are public; the official Hugging Face repository lists Apache 2.0.
Source: Weights and use