Ant Group Unveils Trillion-Parameter AI, Challenging Global Tech Giants

HANGZHOU, China – February 16, 2026 – Ant Group has dramatically raised the stakes in the global artificial intelligence race, releasing a formidable new family of open-source AI models. The announcement includes Ling-2.5-1T, a new flagship trillion-parameter language model, and Ring-2.5-1T, which the company bills as the world's first hybrid linear-architecture thinking model. The release, which also features a powerful multimodal system, signals a significant escalation in the company's capabilities and its ambition to compete directly with established AI leaders like Meta, Google, and OpenAI.

Marking a comprehensive upgrade to its Ling 2.0 series, the new models are now available to researchers and developers worldwide under open licenses on popular platforms Hugging Face and ModelScope. By making these advanced tools publicly accessible under the 'inclusionAI' banner, the fintech giant is not just showcasing its technological prowess but also making a strategic play to shape the future of the open-source AI ecosystem and accelerate its long-term pursuit of Artificial General Intelligence (AGI).

A Leap in Efficiency and Reasoning

At the heart of the announcement is a focus on unprecedented efficiency and advanced reasoning. The flagship model, Ling-2.5-1T, is a powerhouse built on a trillion parameters, but its true innovation may lie in its resourcefulness. The model is designed to deliver top-tier performance with significantly reduced computational cost. For instance, on the AIME 2026 reasoning benchmark, Ling-2.5-1T reportedly matches the performance of frontier models that typically require 15,000 to 23,000 tokens, while using only about 5,890 tokens itself. This dramatic reduction in token usage, combined with support for context lengths up to one million tokens, could translate into substantial cost savings and faster response times for enterprise-scale deployments.

Perhaps more revolutionary is Ring-2.5-1T, a model optimized for complex reasoning. Built on what Ant Group describes as a novel 'hybrid linear architecture,' it employs a sophisticated blend of attention mechanisms to enhance performance on long and intricate tasks. This architecture reportedly yields a tenfold reduction in memory access overhead and triples generation throughput for long sequences, making it ideal for deep thinking and multi-step problem-solving. The company's internal simulations showcase its potential, with Ring-2.5-1T achieving gold-medal-standard scores on benchmarks simulating the International Mathematical Olympiad (IMO) and the China Mathematical Olympiad (CMO). While these are self-tested results, they position Ring-2.5-1T as a state-of-the-art open-source model for mathematics, code, and agent-based tasks. However, such power comes with practical constraints; the model's sheer size would require over a terabyte of memory for deployment, placing it outside the reach of consumer hardware for now.

The Strategic Open-Source Gambit

This release is far more than a technical paper exercise; it is a calculated strategic maneuver in the intensely competitive open-source AI arena. By providing free and open access to models that rival proprietary systems, Ant Group is directly challenging the dominance of Western tech giants and positioning itself as a key architect of the next wave of AI innovation. The choice to brand its open-source efforts as 'inclusionAI' underscores a philosophy of democratizing access to powerful technology, fostering a global community of developers, and establishing its architecture as an industry standard.

This move places Ant Group alongside other major players like Meta with its Llama series and France's Mistral AI, which have also leveraged open-sourcing to drive rapid adoption and innovation. For Chinese technology firms, this strategy is also a way to counter hardware limitations by focusing on algorithmic and architectural breakthroughs. By creating more efficient models, companies can achieve leading performance without relying solely on the most advanced AI chips, a critical advantage amidst ongoing geopolitical trade tensions. The release intensifies the competition, providing developers with more powerful and diverse tools while forcing all players to innovate more quickly.

Beyond Fintech: The Ambitious Path to AGI

Completing the new lineup is Ming-Flash-Omni-2.0, a large-scale multimodal model and the industry's first to unify speech, audio, and music within a single architecture. Positioned as a direct competitor to systems like Google's Gemini, Ming-Flash-Omni-2.0 demonstrates Ant Group's commitment to building AI that can understand and interact with the world through a rich variety of data types.

Together, the Ling (language), Ring (reasoning), and Ming (multimodal) series form a comprehensive AI portfolio that points toward a vision extending far beyond the company's origins in financial technology. Ant Group has explicitly framed these releases as another step in its ongoing journey toward Artificial General Intelligence. This ambition transforms the company from simply an operator of the massive Alipay payments platform into a fundamental research organization aiming to solve one of the biggest challenges in science and technology. The company has stated its belief that AGI should be a 'public good,' and this open-source push is a tangible step toward realizing that vision of creating practical and inclusive AI services.

While the quest for AGI remains a distant goal, the immediate implications for Ant Group's core business are profound. The efficiency gains from Ling-2.5-1T could enable the cost-effective deployment of sophisticated AI features across the Alipay platform, which serves over a billion users. Ring-2.5-1T's advanced reasoning could power a new generation of intelligent financial agents, capable of handling complex user queries and tasks autonomously. Meanwhile, Ming-Flash-Omni-2.0's multimodal capabilities could revolutionize user interfaces, enabling more natural and intuitive interactions through voice and sound. This powerful synergy between foundational AI research and a massive real-world application ecosystem gives Ant Group a unique advantage. As these powerful tools become publicly accessible, Ant Group is not just contributing to the AI community; it is actively reshaping the competitive landscape and accelerating the global journey toward an AI-defined future.