Meet the team that turns your ideas into work.

Discover EightBots
EIGHTBOTS
Contact salesSign up
KyutaiCASA

CASA Helium 1 VL 2B Shared

This variant shares weights between self-attention and cross-attention to connect the Helium text model with visual features.

Research release

The architecture determines how images enter the model

CASA lets text attend to visual features without inserting every image token into the language sequence. The code supplies a custom architecture and processor; this is not an unchanged upstream Qwen or Helium checkpoint.CASA-Helium1-VL-2B-Shared: Kyutai model card

Specifications & access

CASA Helium 1 VL 2B Shared specifications
DetailValue and conditions
StatusResearch release
AccessDownloadable
Model identifierskyutai/CASA-Helium1-VL-2B-Shared
Weight licenseCC BY-NC-SA 4.0.CASA-Helium1-VL-2B-Shared: Kyutai model card
Upstream componentsHelium 1 and a Qwen 2.5 VL encoder. Gemma terms and the Qwen Research License apply.CASA-Helium1-VL-2B-Shared: Kyutai model cardhelium-1-2b: Kyutai model cardCASA-Qwen2_5-VL-3B: Kyutai model card

Related models

  • An instruction-tuned vision-language reference model that inserts image tokens into the text sequence.

    Launch date not confirmedResearch releaseDownloadable