Ant Group Unveils Next-Gen AI Models Ling-2.5-1T and Ring-2.5-1T
Event summary
- Ant Group released Ling-2.5-1T, a trillion-parameter large language model with 1M-token context length and 5,890-token efficiency on AIME 2026.
- Ring-2.5-1T, the world’s first hybrid linear-architecture thinking model, achieved gold-tier results on IMO 2025 (35/42) and CMO 2025 (105/126).
- Both models are open-sourced on Hugging Face and ModelScope under the Ling 2.0 series, first unveiled in October 2025.
- Ant Group also launched Ming-Flash-Omni-2.0, a multimodal model unifying speech, audio, and music in a single architecture.
The big picture
Ant Group’s latest AI model releases reinforce its aggressive push into open-source AI, competing directly with global tech giants. The focus on reasoning efficiency and multimodal capabilities aligns with broader industry trends toward more versatile, scalable AI solutions. With over one billion users connected through Alipay, Ant Group’s AI advancements could further solidify its dominance in China’s digital ecosystem.
What we're watching
- Open-Source Momentum
- How Ant Group’s open-sourcing strategy will impact adoption and competitive positioning in the AI model space.
- AGI Roadmap
- Whether Ant Group’s incremental upgrades signal a viable path toward Artificial General Intelligence (AGI).
- Multimodal Integration
- The pace at which multimodal models like Ming-Flash-Omni-2.0 will disrupt traditional single-modal AI applications.
Related topics