Alibaba · Qwen3.5
Qwen3.5-2B
A 2B dense vision-language model with a 262K-token context, reasoning mode and an official 4-bit MNN build.
- Parameters
- 2B
- Context
- 262,144 tokens
- Input
- Text, Image
- Output
- Text
- Access
- Open weights
- Licence
- Apache 2.0
- Released
- Device evidence
- No device evidence
- Last verified
- Formats
- Safetensors, MNN
- Runtimes
- Transformers, MNN, llama.cpp, MLX
- Platforms
- Android, Linux, macOS, Windows
The official MNN conversion contains a 1.18 GB language weight file plus a 196 MB vision encoder. Compatibility with mobile runtimes is documented, but the release does not include measured phone performance.