MiniMax M3 — NVFP4 NVFP4 quantization of MiniMaxAI/MiniMax M3: 428B parameter (23B active) multimodal MoE with MiniMax Sparse Attention (MSA), 1M token context. 259 GB — runs on 4x 96 GB GPUs at TP4 with headroom for KV cache. What is quantized Component Precision Routed experts — gate proj / up proj / down proj (128 experts x 57 layers) NVFP4 (block 16 E2M1, FP8 E4M3 block scales, FP32 global scale) Dense MLP (layers 0 2) & shared expert down proj NVFP4 Dense MLP (layers 0 2) & shared expert gate up proj BF16 — see note below Attention (q/k/v/o, all 60 layers) BF16 Router ( mlp.gate + e score correction bias ) BF16 / FP32 Embeddings, lm head , all norms BF16 Vision tower + multimodal projector (full VL stack) BF16 — fully preserved The repo is a complete VL model: image/video processors, chat template, and tokenizer ship alongside the weights. Why dense/shared gate up proj is BF16 (an honest note, not hand waving) This is a scar from a calibration config bug, kept and explained rather than papered over. The first export used an over broad exclusion glob gate — meant only for the MoE router ( mlp.gate ) — which silently disabled the quantizers for every gate named module: the route…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy