Ant Group Open-Sources Large AI Model Ling-3.1-flash

Ant Group has made the weights of its Ling-3.1-flash AI model publicly available for developers to deploy on their own infrastructure. The model features approximately 560 billion total parameters, with about 25 billion activated per token, and supports a substantial context window of up to one million tokens.

The release follows the model's earlier introduction and includes deployment support through common inference frameworks such as SGLang and vLLM. It is intended for tasks like coding, agent workflows, and professional applications that involve processing lengthy documents and interacting with external tools.

Ant Group's team has also developed a mixed-precision deployment approach that significantly reduces memory requirements for the model weights, from approximately 534GB down to 292GB. In a reported Hopper GPU configuration, this change reduced the number of required GPUs from eight to four.