InclusionAI Releases Ling-3.1-flash-1019 with 560B Parameters and 1M Context
InclusionAI has introduced Ling-3.1-flash-1019, an upgraded MoE model featuring 560 billion total parameters, 25 billion active parameters per token, and a 1-million-token context window. The model demonstrates significant performance gains across scientific benchmarks and agentic task evaluations. This release marks an important advancement for high-capacity, open-weights mixture-of-experts models designed to handle complex multi-step workflows and long-context processing. Improvements in real-world agentic capabilities make it a strong option for advanced automation tasks. The model achieved an AA-Briefcase score of 1524 Elo, an increase on AutomationBench-AA from 62% to 68%, and higher scores on scientific datasets like CritPt and SciCode. It also registered 44 on the Artificial Analysis Intelligence Index v4.3.2, up from 41 for its predecessor.
## BACKGROUND
Mixture of Experts (MoE) is a neural network architecture where only a subset of parameters, known as active parameters, are used for each input token, allowing for high total capacity with efficient compute usage. A 1-million-token context window enables the model to ingest, analyze, and reason over extremely large documents or codebases simultaneously.