Liquid AI Releases LFM2.5-24B-A2B: A 2B Activated MoE Inference Model for On-Device Agents
Liquid AI has released the open-weight inference model LFM2.5-24B-A2B, featuring 24 billion total parameters with approximately 2.45 billion activated per token and a context window of 131,072 tokens. Designed specifically for multi-step tool-calling agents on laptops and mini PCs, it achieves 184 tok/s on an Apple M5 Max using llama.cpp Q4_K_M quantization.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original