Meituan has unveiled LongCat-2.0, a new large language model developed and trained entirely on domestic semiconductor hardware.
The Chinese delivery and local-services giant stated that the model's performance metrics are comparable to Google's Gemini 3.1 Pro, a benchmark released in February.
The launch underscores Beijing's accelerating efforts to decouple advanced AI development from US-controlled chip supply chains.
The development arrives as Chinese tech firms face persistent restrictions on access to high-end Nvidia GPUs and other advanced US semiconductor equipment.
By demonstrating that domestic chips can support models competitive with leading Western offerings, Meituan aims to validate the viability of China's indigenous AI infrastructure stack.
This narrative is critical for investors assessing the long-term resilience of Chinese tech valuations against geopolitical supply-chain risks.