Meituan has released LongCat-Flash-Lite-Sparse, a new language model built upon LongCat-Flash-Lite. This sparse model replaces dense MLA with LongCat Sparse Attention and supports context lengths up to 1 million tokens, a significant increase from its predecessor's 256k token limit. The model features approximately 3 billion active parameters and utilizes a 30 billion n-gram lookup table offloaded to RAM for efficient handling of long contexts on consumer hardware. AI
IMPACT Enables processing of significantly longer documents and conversations, potentially impacting applications requiring deep context understanding.
RANK_REASON New model release from a significant AI lab (Meituan) with a notable capability improvement (1M context). [lever_c_demoted from frontier_release: ic=2 ai=1.0]
Read on Hugging Face Trending Models →
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →