Trending storyDeveloping
Ant Ling releases Ling-3.1-flash model
1 report1 sourceupdated 1 hour ago
What happened
AI digest
On September 30, 2026, Ant Ling released the Ling-3.1-flash model, upgraded for long real-world tasks. According to AI HOT, it has about 560B total parameters, roughly 25B activated per token, and a context window of up to 1M. The architecture keeps the hybrid linear design but raises the share of linear attention layers: 7 KDA layers to 1 Gated MLA layer, with 8 of 512 routed experts plus 1 shared expert per token.
Written by AI from the coverage · updated 45 minutes ago
Coverage
Follow the reports to see the story from different sides.
Sep 30
- AI HOT · Models蚂蚁百灵发布 Ling-3.1-flash，面向真实世界长任务升级
蚂蚁百灵推出 Ling-3.1-flash，总参数约 560B，每个 Token 激活约 25B，上下文窗口上限 1M。该模型延续混合线性架构并提高线性 Attention 层比例，采用 7 层 KDA 配 1 层 Gated MLA，512 个路由专家选 8 个加 1 个共享专家。
Heat over time
Not enough continuous observations to draw a trend yet.