LFM2.5-24B-A2B: Built for fast on-device agents
Liquid AI released open-weight LFM2.5-24B-A2B, a 24B MoE with 2B active parameters for on-device agents.
Liquid AI released LFM2.5-24B-A2B, an open-weight 24B-parameter mixture-of-experts reasoning model with 2B active parameters, a 128K vocabulary, and a 131,072-token context. It was pretrained on 33T tokens and post-trained with SFT, distillation, and agentic reinforcement learning. It scores 48.09 on LiveClaw versus 38.19 for Qwen3.6-35B-A3B and 27.69 on AutomationBench versus 8.72, while trailing on GPQA Diamond and SWE-Bench Verified. It reaches 184 tokens/s on an Apple M5 Max and up to 21K output tokens/s on one NVIDIA B300.
70