The Zhitong Finance App learned that on October 15, Ant Baling officially opened source Ling-3.1-Flash, allowing the team to test model capabilities and reasoning efficiency in their own tasks and computing power environments. The total parameters of LING-3.1-flash are about 560b, each token is activated about 25B, and supports up to 1M contexts. For local deployment, the team used MXFP4 quantization for Routed Experts, and the density layer maintained block-FP8, reducing the weight usage from about 534GB to 292GB. In the corresponding Hopper GPU test configuration, service deployment was reduced from 8 cards to 4 cards, allowing teams with server-level GPUs to run the model with fewer cards; actual requirements still vary with context, concurrency, and service configuration.
According to reports, when Bering released Ling-3.1-Flash on September 30, it placed the model in scenarios such as office, medical communication, and software development. With this open source, Bering hopes that more teams can deploy and connect to Ling-3.1-Flash in their own environments, incorporate these scenario capabilities into their own business processes, and serve actual needs more flexibly.
According to Artificial Analysis's public evaluation, the intelligence index of LING-3.1-Flash is 41, and the output speed is about 210 tokens per second, providing the dual ability of intelligence index and generation efficiency for high-frequency business scenarios. In Design Arena's Open Weights (Open Weights Model) classification evaluation, Ling-3.1-Flash ranked third in the four directions of data visualization, React Native mobile apps, 3D design, and game development, and ranked fourth and fifth in SVG generation and website production, respectively.
Ling-3.1-Flash can also assist in generating marketing pages, SaaS interfaces, React Native mobile application pages, and editable SVG materials according to text requirements, and can also create Web3D interactive prototypes and lightweight H5 games.