One challenge, every PrismML generation.
Builds the Bonsai family of compressed language and vision models for efficient inference on consumer hardware.
1
1
Sep 2026 to Sep 2026
Publishes the Bonsai family of low-bit models.
Ternary Bonsai 2 27B is derived from Qwen3.8-27B and supports text and images.
Provides model weights and inference tools for NVIDIA GPUs and Apple hardware.