Gemini 3.5 Flash Lands as Google's New Flagship and Goes Generally Available Across the Stack
If you build with Google's models, today is the day to retest your prompts. Gemini 3.5 Flash is now generally available through Google AI Studio, Antigravity, Android Studio, the Gemini Enterprise Agent Platform, the consumer Gemini app, and AI Mode in Search. Google is positioning it as faster, cheaper, and stronger on agentic and coding benchmarks than Gemini 3.1 Pro, with notable jumps on Terminal-Bench 2.1, GDPval and MCP Atlas. If your stack was holding back on Flash because of capability gaps versus Pro, that gap has narrowed enough to revisit the cost trade-off.
Gemini Omni Arrives as a Single Any-to-Any Multimodal Model, Rolling Out to the Gemini App, YouTube Shorts and Flow
Google has folded video generation, image editing and text reasoning into one model rather than stitching them across separate systems. Gemini Omni accepts text, images, audio and video as input and produces consistent multimodal output, with the first variant (Gemini Omni Flash) going live today in the Gemini app, YouTube Shorts and the Flow creative studio. For anyone who currently jumps between a chat model, an image tool and a separate video generator, this collapses the workflow.
It also puts immediate competitive pressure on Runway, Pika and the bespoke video tooling in ChatGPT, because the model is reaching consumers inside an app most people already have on their phone. Subscriber access starts today for Google AI Plus, Pro and Ultra tiers worldwide.
Gemini Spark Turns the Assistant into a 24/7 Personal Agent with Gmail and Calendar Access
Spark is the bigger conceptual shift in today's announcements. Instead of waiting for a prompt, it runs in the background, watches your inbox, calendar and tasks, and acts on your behalf within rails you define. Google is starting it with trusted testers this week and opening the beta to Google AI Ultra subscribers in the United States next week. If you live inside Gmail and Google Calendar, this is the first credible attempt to make Gemini behave less like a chatbot and more like a junior assistant who already has context.
Antigravity 2.0, Managed Agents, Deep Research Max and Google Pics Complete the I/O Picture
Antigravity 2.0 is Google's agent-first development environment, now with a desktop app and a command-line counterpart. A primary agent can spin up specialised subagents inside cross-platform terminal sandboxes, with credential masking and hardened Git policies baked in. Managed Agents go a step further: with a single API call, you get an agent that reasons, calls tools and executes code in an isolated Linux environment powered by Gemini 3.5 Flash, available via the new Interactions API and inside Google AI Studio.
Deep Research Max adds a private data hook so agents can pull from your own corpus alongside the open web, removing the copy-and-paste step that used to leak context. Google Pics is a new design and image generation app inside Workspace aimed squarely at Canva and Anthropic's Claude Design: built in, no second subscription required, and already flagged for tighter Slides and Docs integration. Finally, Google Health Coach launched globally at $9.99 per month, pulling Fitbit, sleep and wellness data into a Gemini-powered coaching layer alongside the new Fitbit Air hardware.
OpenAI and Dell Extend the Codex Partnership to On-Premises and Hybrid Enterprise Data Centres
For enterprises that keep code, documentation and operational systems on-premises for regulatory or sovereignty reasons, this removes a major blocker that previously forced teams to choose between Codex and their existing infrastructure. The plan also covers ChatGPT Enterprise and other API-based products interfacing with Dell systems to prepare data, run tests and deploy applications. If you sit inside a regulated industry that has been holding off on agentic coding tools, this is the route that lets compliance, security and engineering all sign off on the same deployment.
Industry Themes
The whole 24-hour window is effectively a Google I/O window, with the rest of the field deliberately quiet around it. Three patterns are worth flagging. First, the assistant-to-agent transition is no longer a slide: it is a shipping product. Gemini Spark, Daily Brief, Managed Agents and the new Antigravity 2.0 stack are all aimed at the same shift, and OpenAI's Dell deal is the enterprise side of the same story. Second, any-to-any multimodality has moved from research to default, with Gemini Omni collapsing what used to be three or four separate tools into one model and pushing the cost of standalone video generators down. Third, design and creative tooling has become a contested surface for the frontier labs, with Google Pics arriving directly opposite Claude Design and Canva.
All of the Google launches above surfaced first on the company's own blog and the @GeminiApp X account before press coverage caught up, which reinforces the value of treating direct channels as a primary signal rather than a follow-up.