Bfloat16 Python Packages

softfloatpy

A Python binding of Berkeley SoftFloat

mlx-dfloat

Run DFloat11 losslessly compressed BF16 checkpoints on Apple Silicon with MLX and a Metal decode kernel; today that runs FLUX.1 on a 32 GB Mac with weights about 30% smaller

ozaki-jax

FP64-accurate linear algebra on BF16/FP32 hardware (TPU) via the Ozaki scheme, in JAX.

fastflash-attention

Drop-in exact bf16 flash-attention for CUDA with a deterministic backward, tuned for Blackwell (sm_120 / RTX 5090).