Ant Group Open-Sources Ring-1T-Preview, a Trillion-Parameter Reasoning Model
Ant Group has officially released Ring-1T-Preview, marking a significant milestone as the first open-source natural language reasoning model with a trillion parameters. Built upon Ant’s Ling 2.0 foundation, this preview version demonstrates exceptional performance in complex reasoning tasks. In benchmark tests, Ring-1T-Preview achieved a score of 94.69 on CodeForces, surpassing the reported performance of GPT-5. Additionally, it scored 92.6 on the AIME 25 mathematics benchmark, closely trailing GPT-5’s 94.6 while outperforming all known open-source models and Gemini 2.5 Pro. The model also secured the top rank among open-source systems on LiveCodeBench and ARC-AGI-v1, showing partial success in solving IMO25 math competition problems. While the preview version is now available, the full Ring-1T model remains in training. This release highlights Ant Group's advancing capabilities in large-scale artificial intelligence and contributes significantly to the open-source AI community by providing access to a model of unprecedented scale and reasoning proficiency.
Wire timeline
Ant Group Open-Sources Ring-1T-Preview, a Trillion-Parameter Reasoning Model
Ant Group has officially released Ring-1T-Preview, marking a significant milestone as the first open-source natural language reasoning model with a trillion parameters. Built upon Ant’s Ling 2.0 foundation, this preview version demonstrates exceptional performance in complex reasoning tasks. In benchmark tests, Ring-1T-Preview achieved a score of 94.69 on CodeForces, surpassing the reported performance of GPT-5. Additionally, it scored 92.6 on the AIME 25 mathematics benchmark, closely trailing GPT-5’s 94.6 while outperforming all known open-source models and Gemini 2.5 Pro. The model also secured the top rank among open-source systems on LiveCodeBench and ARC-AGI-v1, showing partial success in solving IMO25 math competition problems. While the preview version is now available, the full Ring-1T model remains in training. This release highlights Ant Group's advancing capabilities in large-scale artificial intelligence and contributes significantly to the open-source AI community by providing access to a model of unprecedented scale and reasoning proficiency.
TechNode