Liquid AI released Q4_0 GGUF checkpoints for four LFM2.5 models using Quantization-Aware Distillation (QAD), a technique that distills a high-precision teacher model into a quantized student model. The QAD checkpoints retain roughly 97% of their BF16 baseline performance while preserving the memory efficiency and speed of standard 4-bit quantization. Testing across edge devices including a MacBook Pro, a Raspberry Pi 5, and Samsung Galaxy phones showed the QAD models match or exceed higher-precision alternatives while achieving 3-33% faster decoding throughput.
