Ling 3.1 Flash

Ling 3.1 Flash is InclusionAI / Ant Group's hybrid-reasoning mixture-of-experts flash model, with roughly 560B total parameters and about 25B active per token. It launched API-only, with weights not yet published on Hugging Face, and is positioned for tool-using agent workloads. Hosted context is commonly 262K, while 1M is the announced design target rather than what every gateway serves today. Text input and text output only. The eleven catalog scores are vendor self-reported from the InclusionAI launch benchmark set; there is no first-party or gateway price row stored in this repo yet, so the model is not available through ZeroEval.