TokenRhythm, in collaboration with Wuxinqiong, Tsinghua University, Peking University, and Alibaba Group, has introduced NeoHorse-1, an Agent-Native model. This model, available in 4B and 9B versions, aims to integrate the experiences of agents using tools, receiving feedback, and correcting errors directly into the model's capabilities. NeoHorse-1 builds upon TokenRhythm's previous work with the OpenSquilla routing harness system, extending that research into model training by utilizing agent execution trajectories as core training data. AI
IMPACT This model's training methodology, incorporating agent feedback and error correction, could lead to more robust and adaptable AI agents capable of complex task execution.
RANK_REASON New model release from a lab founded by a former lead of a major LLM project (Huawei Noah's Ark Lab, Pangu Large Model). [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Alibaba Group
- Huawei Noah's Ark Lab
- NeoHorse-1
- On-Policy Distillation
- OpenSquilla
- Pangu Large Model
- Peking University
- Qwen3.5 4B
- Qwen3.5:9b
- TokenRhythm
- Tsinghua University
- Wang Yunheng
- Wuxinqiong
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
