Personal AI agents go viral! Ant releases new model in response, test reveals editorial assistant goes live

What happened: Ant's Bailing released Ling-3.1-flash on September 30 and open sourced it on October 15. The model targets daily office work, long-cycle software development, and complex agent tasks. Key numbers: About 560B total parameters, about 25B activated per token, 1M context window. Why it matters: In Design Arena's Open Weights category, the model ranked third in data visualization, React Native mobile apps, 3D design, and game development. Bailing plans to raise intelligence-per-cost efficiency and lower deployment barriers.

Ant's Bailing released Ling-3.1-flash on September 30 and open sourced it on October 15. The model targets daily office work, long-cycle software development, and complex agent tasks.

In Design Arena's Open Weights category, the model ranked third in data visualization, React Native mobile apps, 3D design, and game development. Bailing plans to raise intelligence-per-cost efficiency and lower deployment barriers.