88Z AI, 6 July — Meituan has officially open-sourced LongCat-2.0, a trillion-parameter large model designed specifically for Agentic Coding tasks. With 1.6T total parameters and 48B average active parameters, the model introduces innovative sparse attention and N-gram Embedding techniques to enhance long-context processing and code execution performance. It has completed inference validation on a domestic computing cluster of 50,000 cards. The release includes multi-precision versions (BF16/FP8/INT8) and optimiz
At a glance
- LongCat-2.0 is designed for Agentic Coding, using sparse attention and N-gram Embedding to boost performance.
- The release includes multi-precision versions and optimized inference code for domestic chips, validated on a 50,000-card domestic cluster.
The update
Meituan has announced the official open-source release of its latest trillion-parameter large model, LongCat-2.0. Designed specifically for Agentic Coding tasks, the model features 1.6T total parameters and 48B average active parameters. Architecturally, LongCat-2.0 innovatively incorporates LongCat sparse attention mechanisms and N-gram Embedding technology, aiming to significantly improve the model's ability to handle long contexts and execute code accurately. Notably, the model has successfully completed full inference validation on a domestic computing cluster comprising 50,000 cards, demonstrating its viability in large-scale domestic hardware environments. This open-source release not only provides model weights in multiple precision formats, including BF16, FP8, and INT8, but also simultaneously publishes inference code optimized specifically for domestic chips. This move aims to lower the barrier for developers to deploy and apply large models on domestic hardware, thereby fostering ecosystem growth.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 龙猫LongCatSource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.