LaunchAI Models12 days agoLiquid AI reveals 8B-A1B MoE model trained on 38T tokensliquid.aiLiquid AI released LFM 2.5, an 8B-A1B mixture-of-experts model trained on 38 trillion tokens. The architecture routes 1B active parameters per input, aiming for efficiency.mixture-of-expertsliquid.ai·via Hacker News·Discuss12 days ago
