Meituan Enters Open-Source AI Race With LongCat Model
Listen to the full version

Chinese delivery giant Meituan has joined the crowded field of open-source artificial intelligence (AI) models, unveiling its own system called LongCat-Flash-Chat Monday across GitHub, Hugging Face and its official website.
The model, nicknamed “LongCat” in Chinese as “龙猫,” uses a Mixture-of-Experts (MoE) architecture with 560 billion total parameters, of which only 18.6 billion to 31.3 billion are activated at a time. While it does not feature reasoning capabilities, Meituan said LongCat-Flash-Chat delivers performance comparable to leading models while activating fewer parameters, excelling especially in agent-based tasks.

Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- DIGEST HUB
- Meituan launched its open-source AI model, LongCat-Flash-Chat, using a Mixture-of-Experts architecture with 560 billion parameters, activating only 18.6–31.3 billion at a time.
- LongCat-Flash-Chat outperformed several mainstream models in agentic tasks but lagged in coding benchmarks; it achieves speeds of 100 tokens/sec and costs $0.69 per million tokens.
- By early 2025, Meituan integrated LongCat into internal tools and customer service, boosting efficiency by 20% in pilot programs.
- Meituan
- Meituan, a Chinese delivery giant, has unveiled its open-source AI model, LongCat-Flash-Chat. It uses a Mixture-of-Experts architecture with 560 billion parameters. The model excels in agent-based tasks and has been deployed internally to power AI tools for coding, meetings, and customer service, boosting efficiency by 20% in pilot programs.
- Alibaba
- Alibaba Group Holding Ltd. launched its Qwen3 model in April and subsequently updated it in July. Its Qwen model was included in benchmark comparisons against Meituan's new AI system, LongCat-Flash-Chat.
- Moonshot AI
- Moonshot AI (月之暗面) is a Chinese tech company that debuted its Kimi-K2 model in July, which quickly topped Hugging Face's global open-source rankings. The Kimi-K2 model was mentioned in benchmark comparisons against Meituan's LongCat-Flash-Chat, where LongCat-Flash-Chat outperformed Kimi-K2 in agentic benchmarks but lagged behind it in coding tasks.
- Baidu Inc.
- According to the article, Baidu Inc. introduced its Wenxin 4.5 model and the inference model X1 in March. This places them among other Chinese tech companies that are actively developing their own large AI models.
- ByteDance Ltd.
- ByteDance Ltd. is a Chinese tech player that has been active in developing large AI models. The company released its Doubao 1.5 reasoning model in April and followed up with version 1.6 in June, showcasing its participation in the competitive AI landscape.
- 2023:
- Meituan co-founder Wang Huiwen pledged $50 million to build 'China’s OpenAI' before stepping back for health reasons.
- By early 2025:
- Meituan had already deployed LongCat internally to power AI tools for coding, meetings, document preparation and short-video production, and rolled out AI-powered customer service and sales agents.
- March 2025:
- Baidu Inc. rolled out its Wenxin 4.5 and inference model X1.
- April 2025:
- Alibaba Group launched Qwen3; ByteDance Ltd. released its Doubao 1.5 reasoning model.
- June 2025:
- ByteDance Ltd. released Doubao 1.6 reasoning model.
- July 2025:
- Alibaba Group updated Qwen3; Moonshot AI debuted Kimi-K2 model.
- Monday, September 1, 2025:
- Meituan unveiled its LongCat-Flash-Chat open-source AI model across GitHub, Hugging Face and its official website.
- PODCAST
- MOST POPULAR