Ant Group unveils Ring-1T, first open-source trillion-parameter reasoning model
Ant Group - the Alibaba affiliate - just spilled some technical details about its latest model, Ring-1T. The company calls it “the first open-source reasoning model with one trillion total parameters,” which would slot it into a very small group of AI projects that have actually hit the trillion-parameter mark. The exact architecture is still under wraps, but the announcement hints that Ring-1T is meant to sit next to reasoning-focused systems like GPT-5 and OpenAI’s o-series.
Open-sourcing something this big isn’t common; it seems Ant’s engineers have found a way around the reinforcement-learning bottlenecks that usually slow down trillion-scale training. If that’s true, developers outside China might finally get a high-capacity reasoning engine without the usual licensing hoops. Still, the real question is how Ring-1T will stack up against its closed-source rivals once the code lands in public repos.
For me, the rollout feels like a clear signal that Ant Group wants to be counted among the top reasoning models shaping today’s AI chatter.
China’s Ant Group, an affiliate of Alibaba, detailed technical information around its new model, Ring-1T, which the company said is “the first open-source reasoning model with one trillion total parameters.” Ring-1T aims to compete with other reasoning models like GPT-5 and the o-series from OpenAI, as well as Google’s Gemini 2.5. With the new release of the latest model, Ant extends the geopolitical debate over who will dominate the AI race: China or the US. Ant Group said Ring-1T is optimized for mathematical and logical problems, code generation and scientific problem-solving.
“With approximately 50 billion activated parameters per token, Ring-1T achieves state-of-the-art performance across multiple challenging benchmarks — despite relying solely on natural language reasoning capabilities,” Ant said in a paper. Ring-1T, which was first released on preview in September, adopts the same architecture as Ling 2.0 and trained on the Ling-1T-base model the company released earlier this month. Ant said this allows the model to support up to 128,000 tokens.
To train a model as large as Ring-1T, researchers had to develop new methods to scale reinforcement learning (RL). New methods of training Ant Group developed three “interconnected innovations” to support the RL and training of Ring-1T, a challenge given the model's size and the typically large compute requirements it entails.
Ant Group says Ring-1T is the first open-source reasoning model that hits a trillion parameters, and it even lists a few engineering tricks to ease reinforcement-learning bottlenecks at that size. The company lines it up next to GPT-5, OpenAI’s o-series and Google’s Gemini 2.5, which sounds like a direct challenge. The catch?
All the numbers come from internal tests - nobody outside the lab has confirmed them yet. Because the code is public, other researchers could take a look, but it also leaves open the question of whether the performance really matches the closed-source rivals. This rollout feeds into the broader China-US AI rivalry that’s been buzzing for years.
If the technical claims survive independent scrutiny, Ring-1T could become a handy benchmark for large-scale reasoning work. On the other hand, without third-party validation its real-world impact stays fuzzy. Either way, Ant Group’s move marks a clear step forward for Chinese AI, even if we’re not sure it will tip the global balance of power.
More in LLMs & Generative AI
Google pushes Gemini rollout to 2026, delays Android Assistant swap
Why does this matter now? For years, Android users have expected a smooth handoff from the familiar...
OpenAI now adds tone controls and streamlined email editing to ChatGPT
OpenAI just rolled out a set of tweaks that feel oddly personal for a chatbot. While the core model...
Google releases FunctionGemma, a tiny model for natural‑language mobile control
Here's the thing: mobile phones are the most common computing platform, yet most AI assistants...
Chinese scammers used AI images for refund fraud; police held buyer eight days
When a shopper in Shanghai posted a video of a defective product, the footage looked...
Gemini pulls Google Calendar data to auto‑suggest meeting times in email replies
Why does this matter? In a flood of AI shortcuts that promise to shave minutes off daily tasks,...
Common Questions Answered
What makes Ring-1T unique as an open-source model according to Ant Group?
Ant Group claims Ring-1T is unique because it is the first open-source reasoning model to achieve a total of one trillion parameters. This places it in an elite category of AI systems and distinguishes it from other models by being publicly accessible at an unprecedented scale.
Which specific AI models does Ant Group position Ring-1T to compete against?
Ant Group positions Ring-1T as a direct competitor to other leading reasoning-focused systems, including GPT-5 and the o-series from OpenAI, as well as Google's Gemini 2.5. This competitive stance highlights the model's ambition to challenge top-tier AI developments primarily from US-based companies.
What engineering achievement does Ant Group highlight for Ring-1T regarding its scale?
Ant Group details that Ring-1T includes engineering fixes designed to overcome reinforcement-learning bottlenecks that occur at the trillion-parameter scale. These improvements are crucial for enabling stable and efficient training and operation of such a massive model.
How does the release of Ring-1T contribute to the geopolitical AI debate?
The release of Ring-1T extends the geopolitical debate over dominance in the AI race between China and the United States. By launching a high-profile, trillion-parameter open-source model, Ant Group, a Chinese company, directly challenges the advancements made by major US tech firms.