FrontierThe story, in brief

LongCat-2.0, a large-scale MoE model with 1.6T total and 48B Active

1.6T parameters. 48B active. LongCat-2.0 just proved MoE scales past the giants.

Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

LongCat-2.0 demonstrates that open-source MoE architectures can achieve competitive scale with frontier models, challenging the narrative that only well-funded labs can build cutting-edge systems. The 48B active parameter count suggests efficient routing and cost-effective inference—a key variable for deployment economics.

The key facts

10 to know
  1. 1.6T total parameters

  2. 48B active parameters

  3. MoE (Mixture of Experts) architecture

  4. Open-source release

  5. Published June 30, 2026

  6. 47 HN points, 12 comments (modest engagement)

  7. LongCat-2.0: 1.6T total parameters

  8. 48B active parameters (MoE architecture)

  9. Hacker News engagement: 47 points, 12 comments

  10. Open-source release via longcat.chat

Go to the source

Hacker Newslongcat.chat

Publisher excerpt: Article URL: Comments URL: Points: 47 # Comments: 12
Read original report
Back to today's editionMore frontier news

Keep reading

Related stories

More from Frontier