AI community optimizes small models and expands local hardware support

The open-source AI community saw various updates in model quantization and local hardware execution. Projects such as Bonsai 2 27B achieved high decode speeds at low bit-widths, while frameworks added day-zero support for new image models.