Model2h ago
LFM2.5 Releases Q4_0 Quantized Checkpoints
The LFM2.5 model now includes Q4_0 checkpoints created via quantization-aware distillation. This process reduces memory overhead while preserving higher precision than standard post-training quantization. Developers can now deploy these weights on consumer hardware with minimal performance loss. It is an incremental update for efficiency, not a leap in raw capability.