Ant Group announced the release and open-sourcing of Ling-1T, a trillion-parameter general-purpose large language model. This launch expands Ant Group’s Ling (also known as BaiLing) model family, which now comprises three main series: the Ling non-thinking models, the Ring thinking models, and the multimodal Ming series.

As a flagship non-thinking model in the Ling family, Ling-1T achieves state-of-the-art (SOTA) performance on multiple complex reasoning benchmarks within constrained output token limits, striking a strong balance between efficient inference and precise reasoning, while delivering improved results across diverse use cases—including code generation, software development, competition-level mathematics problem solving, and logical reasoning.

For example, on the 2025 American Invitational Mathematics Examination (AIME) benchmark, Ling-1T achieves an accuracy of 70.42% at an average cost of over 4,000 output tokens per problem, performing on par with best-in-class AI models.

This follows Ant Group’s release of Ring-1T-preview—the world’s first open-source trillion-parameter thinking model in September.

He Zhengyu, Chief Technology Officer of Ant Group, stated: "At Ant Group, we believe Artificial General Intelligence (AGI) should be a public good—a shared milestone for humanity’s intelligent future. We are dedicated to building practical, inclusive AGI services that benefit everyone, which requires constantly pushing technology forward. The open-source release of Ling-1T and Ring-1T-preview represents a key step in fulfilling our commitment to open and collaborative advancement."

The Ling AI model family now comprises:

l Ling series: Mixture-of-Experts (MoE) non-thinking large language models

l Ring series: Thinking models derived from Ling

l Ming series: Multimodal models processing images, text, audio, and video

l Experimental model: LLaDA-MoE

Together, these models offer diverse sizes and technical capabilities tailored to various application scenarios.