diffusiongemma‑26B‑A4B‑it is a 26B‑parameter (4B active) Gemma 4 MoE block‑diffusion model that generates 256‑token chunks via iterative denoising, delivering up to ~6× higher per‑request throughput than autoregressive Gemma 4 26B‑A4B with a 256K multimodal context window under Apache 2.0.
Getting Started
Step 1
Sign Up and get your Api Key and use with UNLIMITED tokens for 30 days.
Step 2
Paste the URL from Huggingface repository: https://huggingface.co/google/diffusiongemma-26B-A4B-it
Step 3
Choose the GPU machine to deploy.
That’s all! You’re ready to use the model in few minutes without infrastructure complexity in few minutes.
Additional Information
Credits to Google
Applications & Use Cases
- High‑throughput content generation services (summaries, drafts, marketing copy, documentation) where batched requests and total completion time matter more than streaming the first token ASAP.
- Long‑context multimodal analysis over documents plus images (and short videos via frames) using the 256K context window and thinking mode for large‑batch offline processing.
- Function‑calling and agent backends that execute multi‑step reasoning in a single large response, benefiting from the MoE efficiency (4B active) and block‑wise decoding.
- Cost‑sensitive deployments that want near‑Gemma‑4‑31B quality with better GPU utilization and higher tokens‑per‑second per request on a single H100 or similar GPU.
Model details
- Category: Chat
- Provider: Google DeepMind
- Release Date: June 9, 2026
- Total Params: 26B
- License: Apache 2.0