Tag
XiaomiMiMo releases MiMo-V2.5-Pro-FP4-DFlash, an FP4-quantized MoE model with block-diffusion speculative decoding to reduce memory and bandwidth for trillion-parameter inference.
DeepSeek released the full V4 paper detailing FP4 quantization-aware training, MoE training stability tricks (anticipatory routing and SwiGLU clamping), and a generative reward model for RLHF, achieving dramatic efficiency gains—V4-Flash uses only 10% of V3.2's FLOPs and 7% of its KV cache at 1M context length.