From the model card
Fields below are copied from the tags and counters on the HuggingFace repository google/diffusiongemma-26B-A4B-it at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- Pipeline tag
- image-text-to-text
- Library
- Transformers
- Weight formats
- safetensors
- License tag
apache-2.0— read the license file in the repo before relying on it- Downloads (HF counter at last fetch)
- 1,170,594
- Likes (HF counter at last fetch)
- 1,204
- Model card
- https://huggingface.co/google/diffusiongemma-26B-A4B-it
Use cases
- Research into non-autoregressive text generation quality
- Exploring diffusion-based instruction following capabilities
- Multimodal text+image generation research
- Parallel decoding experiments for throughput comparison
Pros
- Apache-2.0 licensed for research and commercial use
- MoE backbone enables efficient diffusion with fewer active parameters
- Multimodal input support for image-conditioned text generation
- Backed by Google research with published architecture details
Cons
- Diffusion text generation requires non-standard inference code incompatible with standard LLM stacks
- Generation quality on open-ended tasks typically trails autoregressive models of comparable scale
- Limited community tooling and ecosystem support compared to AR-based models
- Experimental architecture — not suitable for production deployment without extensive validation
Tags
transformerssafetensorsdiffusion_gemmaimage-text-to-textconversationallicense:apache-2.0endpoints_compatibleregion:us