PrismML releases Bonsai Image 4B, a 930 MB 1-bit diffusion transformer that generates smartphone images in nine seconds
The Apache 2.0 model is based on FLUX.2 Klein.
——0——
QUOTE POST
#604Simo Ryu@CLONEOFSIMO
#604Simo Ryu@CLONEOFSIMOOk but why are they calling quantized flux completely new name? Like why not call it Flux Ternary? Are they following the same "tweak and rename the optimizer" trend? I get they are trying to raise but no mention of flux on this post is disappointing
PrismML@PrismMLToday we’re releasing 1-bit and Ternary Bonsai Image 4B. A new family of image-generation models designed to run high-quality diffusion inference on local hardware: from laptops to phones.
6:21 PM · May 26, 2026 · 39.4K Views
7:29 PM · May 26, 2026 · 1.8K Views
QUOTE POST
#1709Lucas Nestler@CLASHLUKE
#1709Lucas Nestler@CLASHLUKEwe just learned a lot about how Prism’s quantization works
6:46 PM · May 26, 2026 · 1.7K Views