Meituan has launched LongCat-2.5, a multimodal artificial intelligence model with 1.6 trillion parameters, available through an API. It activates approximately 48 billion parameters per token, supports a native context of one million tokens and can generate up to 128,000 tokens. Designed for long-horizon agentic tasks, the model processes images as discrete tokens and was trained with reinforcement learning using failure trajectories and synthetic data. Meituan says LongCat-2.5 rivals several Claude Opus variants on benchmarks for agentic coding and visual tasks. The model was also trained on domestically produced Chinese AI accelerators amid tighter United States restrictions on advanced technologies.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
