This is a quantized version of the inclusionAI Ring-mini-2.0 model. It is quantized using imatrix through llama.cpp and can run more efficiently under different hardware conditions to meet diverse usage needs.

Natural Language ProcessingGguf

Ring-mini-2.0 is a high-performance inference MoE model deeply optimized based on the Ling 2.0 architecture. It has only 16 billion total parameters and 1.4 billion active parameters, yet it achieves comprehensive inference capabilities comparable to dense models below 10 billion scale. It performs excellently in logical reasoning, code generation, and math tasks, supports processing long contexts of up to 128,000 tokens, and enables high-speed generation of over 300 tokens per second.

Natural Language ProcessingSafetensors