Ling 3.2 Flash
by Bailing · family ling-flash · listed Nov 2026RTV
Native multimodal MoE model with 560B total and 25B active parameters for complex visual understanding, multimodal agents, 3D interaction, and medical imaging
Current rates USD per 1M tokens
Input—
Output—
Cache read—
Cache write—
Against the market
On the record
- Model id
Ling-3.2-flash- Context window
- 1M tokens
- Max output
- 33K tokens
- Input modalities
- text, image, video
- Output modalities
- text
- Reasoning
- yes
- Tool calls
- yes
- Open weights
- no
- Knowledge cutoff
- —
- Released
- 2026-11-05
- Last updated
- 2026-11-05
- Provider docs
- developer.ant-ling.com/zh-CN/docs