Discover top fintech news and events!
Subscribe to FinTech Weekly's newsletter
Read by executives at JP Morgan, Coinbase, Blackrock, Klarna and more
Ant Group’s Trillion-Parameter Model Aims to Redefine AI Performance
Ant Group, the Chinese fintech affiliate of Alibaba Group Holding, has announced the release of Ling-1T, a new trillion-parameter large language model (LLM) that the company says outperforms major competitors in code generation, software development, and competition-level mathematics. The company introduced the model on Thursday, describing it as a major step forward in the global race for advanced artificial intelligence.
Ling-1T is Ant’s second LLM with one trillion parameters and is designed as a general-purpose model capable of handling complex reasoning and logic-based tasks. According to Ant, it demonstrated stronger problem-solving skills than both open-source and closed-source rivals from DeepSeek, Moonshot AI, and OpenAI.
Performance Benchmarks and Technical Highlights
Ant reported that Ling-1T achieved higher scores on benchmark platforms such as LiveCodeBench and the American Invitational Mathematics Examination (AIME). The model’s accuracy on AIME reached 70.42 percent, averaging over 4,000 output tokens per problem — results comparable to Google’s Gemini-2.5-Pro and surpassing competing models from DeepSeek, OpenAI, and Moonshot.
The company emphasized that Ling-1T’s advancements stem from optimized training and scaling methods that enable improved reasoning without exponential increases in computational cost. While Ant did not disclose detailed training data or hardware configurations, it described the system as being more efficient in handling multi-step logical processes and large-scale code synthesis.
This performance gives Ant a competitive advantage in areas critical to AI-driven software development, a field where both private and public organizations are rapidly investing.
Part of a Growing Family of AI Models
Ant Group said that Ling-1T joins a broader suite of models under its AI division. These include the Ling series of non-thinking language models, the Ring series of reasoning-based “thinking models,” and the Ming series, which integrates multimodal capabilities for processing images, text, audio, and video.
The fintech company entered the AI model race in 2023 with a financial language model tailored for the digital economy. Since then, Ant has pursued what it calls “practical and inclusive AGI services that benefit everyone.” The introduction of Ling-1T, the company said, represents another milestone in that mission.
Ant previously released Ring-1T-preview, an open-source model it claimed was the world’s first trillion-parameter thinking system. With Ling-1T now positioned as its general-purpose flagship, the company is aiming to push boundaries across sectors including coding automation, mathematics, and data analysis.
China’s Expanding AI Competition
The launch of Ling-1T comes amid an escalating competition among China’s leading AI developers. DeepSeek recently introduced V3.2-Exp, an “experimental” update to its V3 model that improved training and inference efficiency while cutting API costs by more than 50 percent compared to earlier releases. The announcement followed the debut of V3.1-Terminus, which DeepSeek described as its first step toward what it called the “agent era.”
Alibaba, Ant’s parent company, also joined the trillion-parameter club in early September with its Qwen-3-Max-Preview model, designed for large-scale general intelligence applications.
These developments highlight the growing momentum of China’s open-source AI ecosystem, where companies are racing to release models capable of competing with international leaders such as OpenAI, Google, and Anthropic.
A Trillion Parameters and Beyond
Parameters in large language models are adjustable variables that determine how the system learns and generates text. In general, models with more parameters demonstrate more nuanced understanding and contextual reasoning — though they also require vastly greater computational power.
OpenAI’s GPT-4.5, widely regarded as one of the industry’s most powerful models, is estimated to contain between five and seven trillion parameters. While Ling-1T’s trillion-parameter count puts it below that threshold, its open-source nature and specialized reasoning strength make it a strategic move for Ant as it seeks global recognition in the AI sector.
Analysts note that Ant’s push into AI aligns with its fintech foundation, as enhanced reasoning and code-generation tools can streamline digital finance, risk modeling, and automation across its platforms, including Alipay.
Balancing Innovation and Accessibility
Ant Group stated that it aims to build AI technologies that are both advanced and widely accessible. The company said that models like Ling-1T are not just research tools but also practical systems intended to power real-world applications — from enterprise development to scientific computation.
While the firm did not disclose commercialization details, the release signals an intention to compete not only in research benchmarks but also in implementation efficiency and affordability. By open-sourcing such a large model, Ant contributes to a growing global movement that seeks to democratize AI development rather than centralize it among a few major tech companies.
Industry observers in China have remarked that this surge of open innovation may accelerate domestic advancements in critical AI infrastructure, including chips, data centers, and optimization frameworks. If so, Ling-1T could represent both a technological achievement and a strategic signal of confidence in China’s AI capabilities.
Outlook
Ant Group’s Ling-1T underscores how artificial intelligence research is evolving beyond language processing into areas of logic, mathematics, and programming — domains once considered resistant to automation. The company’s claim of superior reasoning performance, if validated by independent testing, could position Ling-1T as a key player in the next phase of global AI competition.
Whether this model will reach the same level of adoption as those from OpenAI or Google remains uncertain. But one thing is clear: the race toward increasingly capable AI systems is accelerating, and Ant Group has firmly placed itself among the contenders shaping that future.
