🤖 Meituan Unveils LongCat-2.0
This is a large-scale MoE model with 1.6 trillion parameters, where approximately 48 billion are active per token. The model supports a context window of 1 million tokens and utilizes LongCat Sparse Attention technology.
🌍 Demonstrates the possibility of efficiently training frontier-scale models on alternative (non-NVIDIA) hardware.
👤 Access to a powerful model for coding and autonomous planning tasks via API in OpenAI and Anthropic formats.
Source 1: https://github.com/meituan-longcat/LongCat-2.0 Source 2: https://longcat.chat/platform/docs/