Ant Group has open-sourced Ling-3.1-flash-1019, an updated version of its Ling-3.1-flash model, along with the base model Ling-3.1-flash-base, both available on Hugging Face under the MIT license.
Ling-3.1-flash-1019 uses a hybrid linear mixture-of-experts architecture with about 560 billion total parameters and roughly 25 billion active per token, with a context window of one million tokens.
On the Artificial Analysis Intelligence Index v4.3.2, the new version scores 44, up three points from the previous release.
Ant says Ling-3.1-flash-base beat DeepSeek-V4-Pro-Base and MiMo-V2.5-Pro-Base on 32 of 47 benchmarks across eight areas, covering knowledge, reasoning, code, maths and long context.