AI Launch Profile
DiffusionGemma
Google DeepMind released DiffusionGemma, an experimental 26B-total-parameter mixture-of-experts open model that generates and refines token blocks through diffusion rather than only sequential next-token decoding.
At a glance
Launch Snapshot
- Company
- Google DeepMind
- Launch date
- June 10, 2026
- Launch type
- Not classified
- Category
- Not classified
- Audience
- AI researchers, model engineers, local-inference developers, and application teams evaluating fast parallel text generation, editing, code infilling, or interactive generation workflows.
- Pricing
- Google DeepMind publishes DiffusionGemma as an experimental open model under the applicable Gemma terms; deployment and hosted inference costs depend on the infrastructure used.
- Free plan
- Not publicly confirmed
- API
- Not publicly confirmed
- Open weights/source
- Not publicly confirmed
Launch Context
Use these links to move from this record into the broader Launch Intelligence database.
Verification & Sources
- Status
- Verified
- Source links
- 4
- Freshness
- Verified July 9, 2026
- Last verified
- July 9, 2026
- Last updated
- July 9, 2026
Key source checks
Suggest a correction
Kingy Launch Score
8.3 / 10 · Solid
One earned credibility score, computed from cited evidence — not a placeholder. How the Kingy Launch Score works
Why this score
- Source & verification 8.5/10 — Official model page, developer guide, Hugging Face model card, and model-card index — four dated first-party sources. deepmind.google
- Product evidence 9.0/10 — Open weights on Hugging Face plus a developer guide — downloadable and directly testable. huggingface.co
- Significance & novelty 8.0/10 — A 26B mixture-of-experts open model generating token blocks through diffusion rather than sequential decoding — technically novel, explicitly experimental. developers.googleblog.com
- Traction signals not scored — insufficient sourced evidence
- Offer clarity 7.0/10 — Published as an experimental open model under the applicable Gemma terms; deployment costs depend on your infrastructure. deepmind.google
Evidence checked 2026-07-10
Kingy AI Take
Google DeepMind released DiffusionGemma, an experimental 26B mixture-of-experts open model that generates and refines token blocks through diffusion instead of purely sequential decoding, with weights and a developer guide published (deepmind.google). AI Researchers, Model Engineers, and Local-Inference Developers can download and test it directly. It is explicitly experimental, so reproduce latency and quality claims on your own hardware before choosing it over autoregressive models.
Who it is for
AI researchers, model engineers, local-inference developers, and application teams evaluating fast parallel text generation, editing, code infilling, or interactive generation workflows.
Source-backed record