Jensen Huang Champions Open Source as Ant Bailing Releases 124B Ling-3.0-flash Execution Model: Beats 1T-Ring-2.6 in 11 of 12 Benchmarks with 12% of the Parameters

TL;DR AI
2 min readKey summary
Ant Group’s Bailing division launched Ling-3.0-flash, a 124B open-source execution model with only 5.1B active parameters.
Despite its compact size, it matched or beat leading models across 34 benchmarks and tied for the top average score.
It also outperformed Ant’s own Ring-2.6-1T by nearly 8 points, underscoring the value of efficiency over sheer scale.
The model is tuned for real-time agent tasks like coding, tool use, and app generation, and supports a 256K context window.
After a free API trial, it will be open-sourced, reflecting China’s continued push toward open AI systems.
