Ling 3.0 Flash
InclusionAI · 127.5B parameters
InclusionAI's MIT-licensed mixture-of-experts model with hybrid reasoning and agentic capabilities. Thinking is enabled by default and can be disabled per request. The complete checkpoint has 127,486,384,608 parameters after excluding 20,992 non-parameter routing-bias buffer elements; the card reports approximately 5.1B active parameters per token.