AlibabaQwen3.5

Qwen3.5-2B

A small vision-language checkpoint intended for prototyping and task adaptation.

Practical use

Practical guidance. Choose a narrow evaluation set that matches the intended fine-tuning task. Report the baseline before training so any improvement can be separated from changes to prompts or data.

Specifications & access

Qwen3.5-2B specifications
DetailValue and conditions
AccessDownloadable
Model identifiersQwen/Qwen3.5-2B
ArchitectureTwo billion parameters; 24 layers.Qwen3.5-2B model card
Context262,144 native tokens.Qwen3.5-2B model card
LicenceApache 2.0.Qwen3.5-2B model card

Related models

  • The large published vision-language checkpoint behind hosted Qwen3.5 Plus.

    Launch date not confirmedDownloadable
  • A mid-sized sparse release in the Qwen3.5 vision-language family.

    Launch date not confirmedDownloadable
  • A dense Qwen3.5 release with image understanding.

    Launch date not confirmedDownloadable