AntLingAGI has announced the release of Ling-3.0-flash, a new hybrid-reasoning model featuring 124 billion parameters and 5.1 billion active parameters, designed to optimize production-scale agent tasks like coding, search, research, and tool integration. This model, which is available free of charge on the Nous Portal for one week, matches or exceeds the performance of the company’s previous flagship model while utilizing a fraction of the parameters.

AntLingAGI: AntLingAGI is an organization developing advanced AI models with a focus on efficient mixture-of-experts architectures optimized for agentic and production workloads. It is actively releasing new models designed to deliver high performance with reduced computational demands. The release of Ling-3.0-flash directly stems from AntLingAGI’s efforts to advance hybrid-reasoning capabilities for real-world agent applications.
Ling-3.0-flash: Ling-3.0-flash is a specific hybrid-reasoning MoE model project created for scalable agent deployment. It is positioned as a lightweight yet capable option that aligns with or surpasses larger flagship models on key benchmarks. The model is being made freely accessible through Nous Portal as part of its launch to encourage immediate testing and integration in coding, search, research, and tool-use scenarios.

Design Focus: The model emphasizes hybrid reasoning tailored specifically to production-scale agent tasks such as coding, search, research, and tool integration.
Release Details: Ling-3.0-flash has been made available at no cost through the Nous Portal for a one-week promotional period following its launch.