Breaking

Google Gemma Simplifies Cloud Run Deployment for DiffusionGemma-Jev Endpoints

September 22, 2026 at 19:14 EDT

Foundation Models
AI Agents
Infra & Chips

Google has made it easier to deploy DiffusionGemma-Jev, or djev, as a Jev API-compatible endpoint on Google Cloud Run with a single command. The update targets developers who want a self-hosted structured-decision service backed by Google DeepMind’s DiffusionGemma rather than a general text-generation chatbot.

Sep 22, 2026 · Google Gemma

DiffusionGemma-Jev gets a Cloud Run path

A single-command deployment now puts a Jev-compatible djev endpoint on Google Cloud Run, aimed at low-latency structured decisions rather than general chat.

35-60 ms

single-step latency

100-123

requests/sec at batch@32

25.2B

total parameters

256K

maximum context length

Latency range, drawn to scale

35 mslow end

60 mshigh end

The Cloud Run endpoint is positioned for fast, bounded decision calls where the answer shape is known before the request is sent.

Why builders care

Typed answers can drive routing, guardrails, scoring, and confidence-gated automation without parsing a generated paragraph.

What still needs proof

Type-safe output does not guarantee correct judgment, and teams still need workflow-specific quality checks before automating decisions.

Deployment story

Cloud Run

serverless GPU container

→

djev

Jev-compatible API

→

DiffusionGemma

structured decision backend

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.