← Back to live feed

Tuesday, Sep 22, 2026

1
Google Launches One Command DiffusionGemma-Jev Deployment on Cloud RunGOOGL
β–Ά
topics πŸ€– AIπŸ’» Tech tags AIAI InfraAI Inference GOOGL keywords Google

Hosting the DiffusionGemma-Jev (djev) model no longer requires researchers or developers to maintain their own GPU infrastructure. Google's new deployment path allows users to spin up API-compatible endpoints on Cloud Run through a single command, with costs of approximately $3 per hour that scale to $0 when the system is idle.

The setup delivers single step latency between 35 and 60 ms and can handle 100 to 123 requests per second for batch sizes of 32. Google released the necessary installation code and instructions on Sept. 22 to streamline experimentation with the tool.

Image via @googlegemma on X
Continued in
vLLM Adds DiffusionGemma-Jev to Ease Cloud AI Deployment
4 tweets β€’ 4 sources
See all 2 tweets β†’