Hello, I'm Masaki Hirokawa from the Gemini Lab editorial team.
The second week of April has wrapped up, and it was packed with significant developments across both the product and technical sides of Gemini. The update that stood out most to me was Google's move to strengthen mental health safeguards within Gemini — a reminder that building responsible AI is just as important as building powerful AI.
Let me walk you through this week's highlights.
Mental Health Support Gets Serious — When AI Takes Responsibility
The most talked-about news this week was Google's expansion of mental health support features in Gemini.
On April 7, Google introduced a new interface that appears when conversations indicate a potential crisis related to suicide or self-harm. It provides one-touch access to helpline resources — call, chat, text, or website — all designed in collaboration with clinical experts. The redesigned "Help is Available" module makes connecting to care more immediate and intuitive.
As AI becomes more capable, it's encouraging to see Google asking not just "what can AI do?" but "what should AI do?" For those of us building with and writing about these tools, being aware of these safety features matters just as much as understanding the technical capabilities.
Android Gemini Gets a Major Visual Overhaul
Starting April 7, Google began rolling out a significant redesign for the Gemini app on Android, and the changes are substantial.
The biggest shift is the new compact overlay — tools and attachments are now consolidated into a single, unobtrusive menu instead of taking over the screen. Gemini Live also received a major update, replacing the large pill-shaped interface with a floating window that shrinks to a small circle when you switch to another app. You can keep your conversation going without it blocking your view.
The whole experience has been updated to Material 3 Expressive, Google's newest design language, with dynamic theming and refreshed iconography. These are the kinds of quiet, thoughtful improvements that make AI feel less like a separate tool and more like a natural part of your phone.
Gemma 4 Open Models — A New Era for Local AI
The Gemma 4 family, released on April 2, continued to generate buzz throughout the week.
Two models — gemma-4-26b-a4b-it and gemma-4-31b-it — are now available through AI Studio and the Gemini API under the Apache 2.0 license. For developers who need local deployments, fine-tuning capabilities, or simply want more control over their AI stack, this is a game-changer.
We published a dedicated guide on Gemini Lab this week covering how to get started with Gemma 4. The expanding open model ecosystem means privacy-focused use cases and edge device AI are becoming increasingly practical.
Gemini API Updates — Flex/Priority Inference and Built-in Tool Combos
Developers had plenty to be excited about this week.
Flex and Priority inference options were introduced, giving developers granular control over the cost-performance tradeoff. Priority inference guarantees resource allocation for low-latency, high-reliability needs, while Flex inference offers a cost-optimized path for batch processing and non-urgent tasks.
Another significant addition: Built-in Tools can now be combined with Function Calling in a single API request. Previously, Google Search, code execution, and other built-in tools had to be used separately from custom function definitions. Now they work together seamlessly, which dramatically simplifies agentic AI application development.
Cloud Storage bucket support and pre-signed URL data input were also launched, with the file size limit jumping from 20MB to 100MB — a meaningful upgrade for RAG systems and document processing pipelines handling large files.
This Week on Gemini Lab
We published a diverse range of articles across all categories this week. Here are some highlights:
API & Development: We focused heavily on troubleshooting guides this week — covering 503 errors, token limit exceeded errors, SDK version mismatches, and Embedding errors. We also published production-oriented guides including Terraform × Gemini API infrastructure automation, and integration guides for PostgreSQL and MongoDB Atlas.
Workspace Integration: New guides on Google Forms × Gemini API automated response analysis and cross-product AI workflows with Apps Script offer practical solutions for business users looking to automate their workflows.
Getting Started: For those newer to Gemini, we published a comprehensive Canvas guide and several Gems custom instruction troubleshooting articles to help you get the most out of these features.
Looking Ahead
With Google I/O 2026 approaching, we're entering the season when teasers for new models and capabilities typically start appearing. The pace of Gemini-related news is likely to accelerate even further.
We're staying on top of the latest developments so we can bring you timely coverage alongside our practical how-to guides. Keep an eye on Gemini Lab next week — there's bound to be more to talk about.
Have a wonderful weekend!