Update Veo Video Endpoints by April 2 to Prevent Service Outages
Google / DeepMind · AI Model Update · · notable
Briefing for: Engineering
What happened
Google is deprecating several 'preview' and 'exp' endpoints for its Veo video generation models, requiring a migration to stable '001' versions by early April. Simultaneously, the new Gemini 3.1 Flash-Lite model has launched in public preview, designed for ultra-low latency and high-volume LLM workloads.
Why it matters
You have a strict 30-day window to update model strings in your production video pipelines to avoid 404 errors. For text-based tasks, the addition of Flash-Lite provides a new optimization target for high-throughput, low-complexity prompts that don't justify the cost of the standard Flash or Pro models.
What this enables
- If you use Veo preview endpoints, you must switch model IDs to the recommended '001' stable versions before April 2.
- If you run high-volume classification or summarization, test Flash-Lite to reduce token costs while improving response speed.
Get personalized AI briefings for your role at Changecast →