Paper 2606.15079
Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale
- Published
- Jun 2026
- Research lab
- Independent
- Citations
- 2
- GitHub
- Not linked
01 In brief
Summary
The report introduces Ling-2.6 and Ring-2.6, a family of trillion-parameter models for efficient agentic intelligence.
Ling-2.6 is optimized for instant responses and high token efficiency, while Ring-2.6 targets deeper reasoning and advanced agentic workflows.
Instead of training from scratch, the models upgrade the Ling-2.0 base via architectural migration pre-training and large-scale post-training.
Key innovations include a hybrid linear attention architecture (Lightning Attention and MLA at a 7:1 ratio) for efficient long-context processing, and post-training methods like Evolutionary Chain-of-Thought, Linguistic Unit Policy Optimization, and bidirectional preference alignment to improve capability per output token.
For agentic capabilities, the report introduces KPop, a reinforcement learning framework using binary KL divergence to stabilize training on environment-grounded data.
Ling-2.6-1T achieves a score of 34 on the Artificial Analysis Intelligence Index using about 16M output tokens, roughly 4x better token efficiency than Ling-2.0-1T.
Ring-2.6-1T achieves 87.60 on PinchBench and 63.82 on ClawEval.
The models are open-sourced, and the report details pre-training, post-training, infrastructure, and evaluation results.
02 From the paper
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.