Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale

2026-06-12

AI Summary

This work addresses the fundamental trade-off between low-latency responsiveness and strong reasoning capabilities in trillion-parameter agent systems by introducing the Ling-2.6 (for immediate response) and Ring-2.6 (for deep reasoning) model families. Through co-optimization of model architecture, training objectives, and serving infrastructure, the proposed approach integrates hybrid linear attention (combining Lightning Attention and MLA), Evolutionary Chain-of-Thought reasoning, bidirectional preference alignment, shortest-correct-response distillation, and the KPop asynchronous reinforcement learning framework. These innovations substantially enhance long-context processing efficiency, per-token reasoning capacity, and agent training stability. The project successfully constructs a highly efficient and scalable trillion-parameter agent system and publicly releases all checkpoints of the 2.6 series models.

Abstract

Efficient and scalable agentic intelligence requires models that can deliver both low-latency responses and strong reasoning capabilities while remaining practical to train, serve, and deploy. In this report, we present Ling-2.6 and Ring-2.6, a family of models designed to address this challenge at scale. Ling-2.6 is optimized for instant response generation and high capability per output token, whereas Ring-2.6 is tailored for deeper reasoning and more advanced agentic workflows. Instead of training from scratch, we upgrade the Ling-2.0 base model through architectural migration pre-training and large-scale post-training. This upgrade is guided by a unified co-design of model architecture, optimization objectives, serving systems, and agent training environments, enabling improvements in both model capability and deployment efficiency. At the architectural level, we introduce a hybrid linear attention design that integrates Lightning Attention with MLA, improving the efficiency of long-context training and decoding. To further enhance token efficiency, we optimize capability per output token through Evolutionary Chain-of-Thought, Linguistic Unit Policy Optimization, bidirectional preference alignment, and shortest-correct-response distillation. For agentic capabilities, we propose KPop, a reinforcement learning framework designed to support stable training of Ring-2.6-1T on large-scale environment-grounded data. KPop improves training efficiency through asynchronous scheduling across coding, search, tool use, and workflow execution, enabling scalable learning from complex agent-environment interactions. Together, Ling-2.6 and Ring-2.6 provide a practical pathway toward efficient, scalable, and open agentic systems. We open-source all checkpoints in the 2.6 family to support further research and development in practical agentic intelligence.

Problem

• agentic intelligence
• low-latency response
• scalable reasoning
• trillion-parameter models
• efficient deployment

Innovation

• hybrid linear attention
• Evolutionary Chain-of-Thought
• KPop
• asynchronous scheduling
• agentic intelligence