Ant group unveils Ling-3.1-flash model with 560b parameters

On September 30, 2026, Ant Group's Bai Ling announced the release of the Ling-3.1-flash model, featuring approximately 560 billion total parameters and 25 billion active parameters per token. The model supports a context window of up to 1 million tokens, enabling it to handle longer documents, code, and task histories. Ant Group emphasized ongoing improvements in areas such as general-purpose agents, search, daily office tasks, and software development, with enhanced capabilities in medical, financial, and materials science research. The Ling-3.1-flash model will offer a two-week free trial, during which the service limit is set at 256K tokens. After the trial period, the model will transition to a paid service, with plans to expand the context window to 1 million tokens and simultaneously open-source the model, with continued performance updates.