meituan-longcat/LongCat-2.0

meituan-longcat/LongCat-2.0

License: MIT

Stars: 41

Forks: 2

Open issues: 0

Created: 2026-06-29T12:46:02Z

Pushed: 2026-06-30T02:43:25Z

Default branch: dev/longcat-preview

Fork: no

Archived: no

README:

LongCat-2.0

Tech Blog 📄

Model Introduction

We introduce LongCat-2.0, a large-scale MoE language model with 1.6 trillion total parameters and ~48 billion activated per token — a substantial step up from previous LongCat models, accompanied by several architectural improvements.

Both the full training run and the large-scale deployment are built entirely on AI ASIC superpods. Pretraining spans millions of accelerator-hours across more than 35 trillion tokens, with no rollbacks or irrecoverable loss spikes — demonstrating that we have the capability to conduct frontier-scale training on alternative hardware platforms.

To strengthen the model on long-horizon tasks, we introduce LongCat Sparse Attention and train LongCat-2.0 on hundreds of billions of tokens of 1M-context data. Together with dedicated post-training, this gives LongCat-2.0 strong performance on coding and agentic tasks.

---

> [!NOTE] > 🏋️ Model weights coming soon — stay tuned!