Post-Transformer Pioneer Liquid AI Unveils LFM 2.5: The True Value of Next-Gen Architecture Demonstrated by an 8B-MoE Model Trained on 38 Trillion Tokens
Liquid AI has officially released LFM 2.5-8B-A1B, an 8.3B parameter Mixture-of-Experts model trained on a massive 38 trillion tokens. By utilizing a 'Post-Transformer' architecture, the model achieves frontier-level reasoning and a 128K context window while remaining small enough to run on consumer laptops and smartphones.