GEMINI LABJP
NANOLITE — Nano Banana 2 Lite is here: Google's fastest and most cost-efficient Gemini Image model, made for running lightweight image generation cheaplyOMNIFLASH — Gemini Omni Flash is in public preview, a natively multimodal model that lets enterprises and developers build custom, dynamic video workflowsAGENTS — Managed Agents expand with background: true for async server-side runs and polling, remote MCP server integration, and refreshing credentials across interactionsMEMORY — The Memory Bank IngestEvents API is generally available, decoupling event ingestion from memory generation so you can stream content continuouslyTHROUGHPUT — Provisioned Throughput now lets you submit up to seven pending orders for the same model and regionDEPRECATE — Image generation models shut down on August 17, and the Grok 4.1 family on the Gemini Enterprise Agent Platform on August 20NANOLITE — Nano Banana 2 Lite is here: Google's fastest and most cost-efficient Gemini Image model, made for running lightweight image generation cheaplyOMNIFLASH — Gemini Omni Flash is in public preview, a natively multimodal model that lets enterprises and developers build custom, dynamic video workflowsAGENTS — Managed Agents expand with background: true for async server-side runs and polling, remote MCP server integration, and refreshing credentials across interactionsMEMORY — The Memory Bank IngestEvents API is generally available, decoupling event ingestion from memory generation so you can stream content continuouslyTHROUGHPUT — Provisioned Throughput now lets you submit up to seven pending orders for the same model and regionDEPRECATE — Image generation models shut down on August 17, and the Grok 4.1 family on the Gemini Enterprise Agent Platform on August 20
Articles/Updates
Updates/2026-06-11Beginner

Google I/O 2026 Preview — What I'm Watching for in Gemini This Year

Google I/O 2026 is approaching. Based on current Gemini development trends and past announcement patterns, here's what I'm personally expecting — no guarantees.

Google I/O 20263Gemini75Google AI14Gemini 35latest newspreview

Google I/O 2026 is almost here. Every year in mid-May, this is where Google puts its biggest announcements on stage — and for the past few years, that's meant major Gemini news.

Last year's I/O brought Gemini 2.5 and a significant expansion of Project Astra. It surprised me more than I expected. So what's coming this time? Here's what I'm watching.

The Gemini 3 Model Family Taking Shape

The Gemini 3 series has been rolling out steadily through early 2026 — Gemini 3 Flash, 3 Pro, 3.1 Flash, 3.1 Pro, with 3.2 in limited preview for certain features. At this point, the lineup feels somewhat scattered.

I'm expecting I/O to present a clear, organized picture of the full Gemini 3 family: how the models relate to each other, which plan gives you access to which model, and what each is actually best for.

What I'm most curious about is Gemini 3 Ultra. With Gemini 2.5, Ultra was in limited access. If I/O opens it fully to Google AI Pro and Ultra subscribers, that would meaningfully change the daily experience for power users.

Gemini Nano Growing Up

Gemini Nano — the on-device model — has been spreading across Google's device ecosystem since late 2025. Pixel 9 integration, Chrome's Prompt API, Android-wide rollout. The foundation is there.

I'm expecting I/O to show the next phase: richer multimodal capabilities running on-device (offline image analysis, more capable voice understanding), and possibly more detail on the Apple partnership announced last year and what it means for Gemini Nano on iPhones.

Workspace Integration Getting Smoother

Gemini in Workspace has improved considerably — Gmail reply suggestions, Docs drafting, Sheets data analysis are all genuinely useful now. But the rough edges are still visible when you try to work across products.

What I'd like to see is cross-product context awareness: something like "take the notes from last week's Meet recording and build this week's Slides deck from them" working with minimal prompting. Currently that kind of multi-product flow requires too many manual steps. If I/O addresses this, Workspace Gemini becomes meaningfully more powerful for everyday use.

Project Astra Going Mainstream

Project Astra has been in limited access for a while now. The demos — real-time vision combined with voice conversation — look genuinely useful. It feels like it's ready for a broader release.

My question is Japanese-language performance. Early Astra demos have been primarily English-first. If I/O shows a version that works well in Japanese for everyday use, that's a significant shift for how I'd use it.

Veo and Video Generation's Next Move

Veo 3 and Veo 3.1 Lite are already available through the API and Google AI Studio. The quality is good enough for product use cases now.

I expect I/O to show more creative production workflows — expanding Google Flow, better pricing transparency for API usage, and possibly new features like improved image-to-video and longer generation limits. Video generation going from "interesting experiment" to "practical production tool" feels like the trajectory.

What I'm Actually Most Interested In

Honestly, I'm less interested in benchmarks and more interested in whether Gemini gets easier to use for complex workflows.

2025 was a year of capability leaps. 2026 feels like it should be about usability. Getting multiple Gems to work together, maintaining a consistent voice across long documents, using Deep Research output as input for other tasks — these are all still clumsier than they should be.

Last year, the NotebookLM project management feature quietly changed how I do research and writing. I'm hoping for something similar this year — a feature that changes how I work without me necessarily expecting it beforehand.

How to Watch I/O

Google streams the keynote live on YouTube. The timing works out to late night or early morning for Japan. The full keynote archive is always available immediately after, so there's no pressure to stay up if you'd rather watch it fresh in the morning.

I'll post a summary and initial hands-on impressions here after the announcements. What I'm actually curious about is the gap between what gets announced and what works well in daily Japanese-language use — that's usually where the real story is.


June 2026 update: what actually shipped at I/O 2026

Google I/O 2026 took place on May 19, 2026. I said ahead of time that ease of use was what I would watch most closely; in the end the announcements centered on stronger models and a more coherent agent experience. Here are the Gemini items worth holding onto, lined up against the predictions above.

  • Gemini 3.5 Flash: the first of the new series. It delivers intelligence rivaling the large flagship models at Flash-series speed, and reportedly beats Gemini 3.1 Pro on coding and agentic benchmarks, running long-horizon autonomous work at less than half the cost of other frontier models.
  • Gemini 3.5 Pro: already in internal use, with general rollout flagged for the following month.
  • Gemini Omni / Omni Flash: a model that creates any output from any input, starting with video. Omni Flash is rolling out across the AI Plus, Pro, and Ultra tiers through the Gemini app and Google Flow.
  • Managed Agents (Gemini API): a single API call now provisions an isolated Linux environment where an agent can reason, plan, call tools, and manage files.
  • Antigravity 2.0 and the Antigravity CLI: shipped as an agent-first development platform where you spin up specialized subagents to handle complex workflows.

The "seams between Gems and Deep Research" I hoped to see smoothed over moved forward from a slightly different angle, through Managed Agents and the Omni models. How much that actually changes everyday Japanese-language work is something I will test hands-on and report back on.

Share

Thank You for Reading

Gemini Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • Copy-paste ready implementation code
  • New advanced guides published daily
  • $5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

Updates2026-04-09
Gemini Updates — April 2026 Roundup: Gemini 3, Voice, Workspace, and AI Studio
A roundup of the most important Gemini and Google AI updates in April 2026: Gemini 3 rollout, voice interface improvements, Workspace feature additions, AI Studio UI changes, and safety policy updates.
Updates2026-04-06
Google AI News Roundup: April 2026 — Gemini Updates and What's New
A complete roundup of Google AI and Gemini updates for April 2026 — model improvements, Google AI Studio features, Gemini Code Assist enhancements, Workspace AI integration, and developer API changes.
Updates2026-03-18
Google AI March 2026 Roundup — Groundsource, Gemini Workspace Revolution, Flash-Lite, and Pixel Drop
Comprehensive overview of Google AI developments in March 2026. Disaster prediction with Groundsource, Gemini Workspace revolution, Gemini 3.1 Flash-Lite Preview, Pixel Drop features, and Chrome expansion.
📚RECOMMENDED BOOKS
Build a Large Language Model (From Scratch)
Sebastian Raschka
LLM Dev
Prompt Engineering for LLMs
Berryman & Ziegler
Prompting
AI Engineering
Chip Huyen
AI Eng
* Contains affiliate links
See all →