Nanbeige4.1-3B
Nanbeige LLM Lab · 2026-02-10 · 3.9B parameters
Nanbeige4.1-3B is Nanbeige LLM Lab's 3B unified generalist model released in February 2026, built on the Nanbeige4-3B-Base with a post-training pipeline combining large-scale supervised fine-tuning and multi-stage reinforcement learning for simultaneous reasoning, preference alignment, and agentic capability. It is notable as one of the first open small language models to natively support deep-search workflows involving sustained multi-round tool use. The accompanying technical paper is titled "A Small General Model that Reasons, Aligns, and Acts."