Ant Group Releases Its First Open-Source Trillion-Parameter Inference Model, Ascending to the 1T Model Echelon

1 week ago / Read about 0 minute

Author：小编

In the wee hours of the morning, Ant Group made a groundbreaking move by open-sourcing its first self-developed trillion-parameter large model, dubbed Ring-1T-preview. This milestone marks the debut of the world's inaugural open-source trillion-parameter inference large model. The preview iteration of the model showcases remarkable natural language reasoning prowess, excelling in tests and authoritative benchmarks like AIME 25 and CodeForces, while also demonstrating commendable performance in the IMO25 test. The Ant Bailing team is unwaveringly committed to advancing the post-training endeavors for the 1T language foundation model within the Ling2.0 family, with the official version of Ring-1T currently undergoing training.

Previous page：Microsoft Unveils Restructuring of Windows Team, M...

Next page：Jensen Huang Makes a Daring Forecast: OpenAI Could...

Return to List

Hot Reading

1 day ago

OpenAI, Jony Ive's AI-Powered Device Could be Delayed Due to 'Technical Issues,' Claims Report

2 day ago

The Reinforcement Gap — or why some AI skills improve faster than others

2 day ago

OpenAI and Jony Ive may be struggling to figure out their AI device

1 day ago

Apple Wallet's iOS 26 Boarding Passes Now Supported by United Airlines