[HUGGINGFACE]score: 0.46
New model: bonsai-image-ternary-4B-gemlite-2bit by prism-ml
June 1, 2026
A 4B-parameter vision model quantized to 2-bit ternary weights using the GemLite format, released by prism-ml. The ternary quantization scheme reduces memory footprint significantly compared to standard INT4 or FP16 deployments while targeting image understanding tasks.
diffuserssafetensorsternary1.58-bitgemlitehqqcudatext-to-imagediffusionfluxprismmlbonsaibase_model:prism-ml/bonsai-image-ternary-4B-unpackedbase_model:finetune:prism-ml/bonsai-image-ternary-4B-unpackedlicense:apache-2.0diffusers:Flux2KleinPipelineregion:us