Meituan Open Sources LongCat-2.0 Below MIT License – Open Supply For You


Meituan logo
Meituan brand

Meituan has open sourced its 1.6T-parameter LongCat-2.0 AI mannequin below the MIT License, giving builders and enterprises a commercially pleasant agentic coding mannequin with a 1-million-token context window.

Meituan has open sourced LongCat-2.0, releasing the 1.6-trillion-parameter Combination-of-Consultants (MoE) mannequin on GitHub, Hugging Face and its personal platform below the permissive MIT License. Beforehand generally known as the nameless Owl Alpha mannequin, LongCat-2.0 topped OpenRouter developer rankings earlier than its id was revealed.

The enterprise-friendly MIT License permits builders to switch, redistribute and combine the mannequin into proprietary software program with out being required to open supply spinoff works, making it one of many yr’s most commercially versatile open-source AI releases.

Designed for autonomous coding, software program engineering, AI brokers and repository administration, LongCat-2.0 helps a local 1-million-token context window. It was skilled fully on a cluster of greater than 50,000 Chinese language ASIC chips, demonstrating that near-frontier AI fashions might be developed with out counting on Nvidia GPUs.

Earlier than its public unveiling, Owl Alpha processed round 10.1 trillion month-to-month tokens, averaging 559 billion tokens each day, whereas recording 242% month-on-month progress. It ranked amongst OpenRouter’s world prime three fashions and secured the highest spot on Hermes Agent Workspace.

LongCat-2.0 scored 59.5 on SWE-bench Professional, narrowly outperforming GPT-5.5’s 58.6, underscoring its energy in software program engineering. The mannequin additionally introduces LongCat Sparse Consideration, Zero-Compute Consultants and the MOPD post-training framework to enhance long-context effectivity and agent efficiency.

Meituan says organisations can self-host the mannequin to cut back API dependence, enhance information privateness and decrease operational prices by means of free context-cache hits, positioning the corporate as an rising world open-source AI infrastructure supplier.