Google has released DiffusionGemma, an open-weight generative AI model under Apache 2.0 license, representing the return of their experimental Gemini Diffusion research. The 26B-parameter model is now available via NVIDIA's NIM cloud API, demonstrating impressive performance of over 500 tokens per second in initial tests. This release makes advanced text-to-image generation more accessible to developers and researchers.
Background
Google previously experimented with Gemini Diffusion models but made no major announcements until now. The release of DiffusionGemma as an open-weight model represents a significant step in making advanced AI research more accessible to the broader community.
- Source
- Simon Willison
- Published
- Jun 11, 2026 at 04:00 AM
- Score
- 7.0 / 10