Meituan Open-sources 1.6 Trillion-Parameter LongCat-2.0
According to public information, Meituan has open-sourced the 1.6 trillion-parameter LongCat-2.0 model. The training process utilized 5 万 domestic acceleration cards, and the model is released under the MIT license.
Meanwhile, research institutions also released lightweight models optimized for edge devices, further reducing the cost of edge inference. Industry experts believe that domestic computing power successfully completing engineering validation for ultra-large-scale model training serves as a meaningful benchmark for boosting confidence across the supply chain.
Source: Compiled from public technical communities and media reports.
Karpathy introduces the "Capability Cognition Funnel."
NextOpenAI fires three researchers amid internal security debate
Related Articles
Domestic First AI Application Risk Comprehensive Assurance Solution Goes Live
Mitigate liability for infringing AI-generated content and performance deviations with a comprehensive risk protection solution.
US Big Four Banks Activate AI Agent Fund Operation Contingency Plan
Define a response workflow for anomalous agent account operations.
Linux Foundation Report: AI Coding Increases Open Source Maintenance Burden
The report highlights new maintenance and governance challenges posed by AI-generated code.