Run fast, private agents on-device with LFM2.5-24B-A2B

This post introduces LFM2.5-24B-A2B, a 24B-parameter reasoning model with 2B active parameters, designed for fast, private, on-device agents. It details the model's competitive performance against larger models on agentic tasks (scoring 48.09 on LiveClaw vs. 38.19 for Qwen3.6-35B-A3B) and its efficient inference speed (184 tok/s on an unspecified GPU). The post emphasizes the model's ability to think before answering and call tools, making it suitable for continuous agent use on local hardware.