Bonsai Image
#5 in AI Image Editingprismml · since 2026-05-26 · 7× · last seen Jun 30, 2026
45
Momentum
Bonsai Image 4B is a family of heavily compressed text-to-image diffusion models by PrismML, released on May 26, 2026, built on top of FLUX.2 Klein 4B. It comes in two variants: 1-bit (binary, 0.93 GB transformer) and ternary ({-1, 0, +1}, 1.21 GB transformer). Both variants are designed for local inference on Apple Silicon devices (iPhone, iPad, Mac) and NVIDIA CUDA GPUs, and are released as open weights under the Apache 2.0 license at no cost. Weights and code are available on Hugging Face.
Momentum trend
04.04.03.07.
Features
| Local Executability | Fully runnable locally on Apple Silicon iPhone, iPad, Mac (via MLX) and NVIDIA CUDA GPUs on Linux and Windows (via Gemlite); first 4B-class image model to run directly on iPhone |
| Maximum Resolution | 1024×1024 px (native design target resolution); aspect ratios in multiples of 32 supported (e.g., 832×1248, 1248×832); 512×512 for quick previews |
| Entry Price | Free – open weights and code under Apache 2.0 license (no commercial API pricing published) |
| Language Support | Natural-language text prompts (English); no multilingual support stated in official documentation; negative prompts not required |