LFM2.5-24B-A2B on-device reasoning agent announced with expanded context and speed

Hugging Face announced LFM2.5-24B-A2B, a fast, private on-device agent model 1. The model scores 48.09 on LiveClaw, outperforming Qwen3.6-35B-A3B's 38.19 on the same benchmark 1. It uses a Mixture-of-Experts architecture with 40 layers, 64 experts per block, top‑4 routing and only 2 B active parameters per token, making reasoning tokens cheap 1. Inference runs at 184 tokens per second on an Apple M5 Max and up to 21 000 output tokens per second on a single NVIDIA B300 GPU 1. The context window expands to 131 K tokens and the vocabulary doubles to 128 K tokens, while the model generates an explicit chain of thought before answering; training scaled to ~33 T tokens with additional 2 T and 300 B token phases focused on reasoning, code, multilingual text and long‑context data, followed by five post‑training stages to become an agent 1.

Why it matters. The model can run fast, private reasoning directly on devices, enabling on-device use 1. Its larger context window and doubled vocabulary improve handling of long texts and non‑Latin scripts 1.