Meituan Unveils Record-Breaking Open-Source LLM
Meituan has released LongCat-2.0, a 1.6-trillion-parameter open-source large language model, trained on over 50,000 domestic AI accelerators without Nvidia hardware. This milestone comes amid China's expansion of domestic computing capabilities, amid export restrictions on US graphics processors. The model supports a 1-million-token context window, rivaling DeepSeek's V4-pro.
Key points
- Meituan, a Chinese food delivery giant, has unveiled LongCat-2.0, a 1.6-trillion-parameter open-source large language model.
- The model was trained on over 50,000 domestic AI accelerators, making it the first trillion-parameter model to achieve this scale without Nvidia hardware.
- LongCat-2.0 supports a 1-million-token context window, rivaling DeepSeek's V4-pro, which launched in April.
- The announcement comes amid China's expansion of domestic computing capabilities, amid export restrictions on US graphics processors.
- Analysts say this milestone highlights China's growing capabilities in AI and computing, despite export restrictions.
- The model's release is expected to contribute to the development of China's AI ecosystem and potentially rival global leaders in the field.
Meituan Unveils Record-Breaking Open-Source LLM
Meituan, a Chinese food delivery giant, has made a significant breakthrough in the field of artificial intelligence (AI) by releasing LongCat-2.0, a 1.6-trillion-parameter open-source large language model. This achievement is notable not only for its sheer scale but also for the fact that it was trained on over 50,000 domestic AI accelerators without relying on Nvidia hardware.
The model's capabilities are impressive, supporting a 1-million-token context window, which puts it on par with DeepSeek's V4-pro, a flagship model launched in April. This milestone comes amid China's expansion of domestic computing capabilities, amid export restrictions on US graphics processors.
Analysts say this achievement highlights China's growing capabilities in AI and computing, despite export restrictions. The release of LongCat-2.0 is expected to contribute to the development of China's AI ecosystem and potentially rival global leaders in the field.
The model's training process was a significant undertaking, requiring a computing cluster containing more than 50,000 domestic AI accelerators. Meituan has credited Huawei's Collective Communication Library for improving communication stability across processors.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.