# avifenesh/bw24

From-scratch Rust+CUDA inference engine, bit-exact by construction — NVFP4, MoE, MTP speculative decoding, tuned against measured limits of one RTX 5090 Laptop (sm_120a).

**Commercial license**：Commercial OK — 可商用，通常只需保留著作權聲明/授權條款

**Stars**：278
**Source**：https://github.com/avifenesh/bw24