Vascend
Horizon
Vascend IDE
Vascend CLI
Community
Benchmark
Contact
EN
EN
Sign in
Join Discord
News
Videos
Courses
Stack
Projects
Discord
News
Videos
Courses
Stack
Projects
Apple
All
Models
AI Models
Oct 14, 2026
LFM2.5-24B-A2B: reasoning MoE per agent in locale
Liquid AI rilascia LFM2.5-24B-A2B, il più grande della famiglia LFM2.5: 24B parametri totali ma solo 2B attivi per token, contesto da 131K e vocabolario da 128K.
Join the community
Developers, founders and AI enthusiasts. Free to join.
Join the Discord