Ant Group's Ling AI: Trillion-Parameter Model

Ant Group Unveils Ling AI Model Family with Ling-1T

Ant Group, the Chinese fintech giant, has officially launched its Ling AI model family, highlighted by the release and open-sourcing of Ling-1T, a trillion-parameter large language model (LLM). This milestone, announced in October 2025, marks a significant advancement in AI technology, positioning Ant Group among global leaders in large-scale AI model development.

A New Era for Ant Group’s AI Strategy

Ant Group’s Ling AI model family, also known as BaiLing, comprises three core series tailored for diverse AI applications:

Ling series: Non-thinking models focused on efficient inference and precise reasoning.
Ring series: Thinking models, including the recently released open-source Ring-1T-preview.
Ming series: Multimodal models designed for cross-modal understanding and generation.

The release of Ling-1T expands this ecosystem as the flagship non-thinking model, boasting one trillion parameters, a scale that rivals or surpasses many existing large language models globally.

Ling-1T’s Technical Achievements

Ling-1T is engineered to deliver state-of-the-art (SOTA) performance across complex reasoning tasks while maintaining efficiency in inference. One of the standout achievements is its performance on the 2025 American Invitational Mathematics Examination (AIME) benchmark, where Ling-1T attained an impressive 70.42% accuracy with an average output token cost of over 4,000 tokens per problem. This level of performance places Ling-1T on par with the best AI models available internationally for high-level mathematical problem-solving.

Beyond mathematics, Ling-1T demonstrates strong capabilities in:

Code generation and software development
Logical reasoning tasks
General-purpose language understanding and generation

According to Ant Group, the model strikes a balance between efficient inference and precise reasoning, making it highly versatile for industrial and research applications.

Open Source and Industry Impact

The open-sourcing of Ling-1T reflects Ant Group’s commitment to contributing to the global AI community. This move allows researchers and developers worldwide to access and build upon a trillion-parameter model, which could accelerate innovation in natural language processing (NLP) and AI-driven services.

This development follows Ant Group’s earlier launch of the Ring-1T-preview in September 2025, which was the world’s first open-source trillion-parameter "thinking" model. Together, these models form a comprehensive AI framework capable of addressing a wide array of tasks from reasoning to multimodal understanding.

Strategic Implications and Future Prospects

Ant Group’s push into trillion-parameter models signifies a strategic investment in AI that aligns with China’s broader ambition to lead in artificial intelligence technology. The Ling family’s combination of non-thinking, thinking, and multimodal models provides a robust platform for:

Enhancing fintech products, including fraud detection, credit risk assessment, and customer service automation.
Expanding into AI-assisted software development and scientific research.
Driving innovation in AI-powered education and competition-level problem solving.

Given Ant Group’s established position in digital payments and financial services, the integration of Ling AI models could lead to substantial improvements in product intelligence and user experience.

Visuals Related to the Launch

Relevant visuals to accompany this story include:

The official Ant Group logo and branding.
Diagrams or screenshots illustrating the Ling AI model architecture and series.
Performance charts showcasing Ling-1T’s benchmark results, such as the AIME performance comparison.
Photos from the official launch event or Ant Group’s research facilities in Hangzhou, China.

Ant Group’s unveiling of the Ling AI family and the trillion-parameter Ling-1T model underscores the rapid evolution of AI technology in China and worldwide. By open-sourcing Ling-1T, Ant Group not only enhances its own technological ecosystem but also contributes significantly to the democratization and advancement of large language models in the global AI landscape.