Gemini 3.5 Flash
"Frontier intelligence with action." 4× faster than other frontier models at less than half the price; built for agentic workloads.
Everything new, distilled. May 19–20 · opening + developer keynote · 33 announcements across 8 categories.
"Frontier intelligence with action." 4× faster than other frontier models at less than half the price; built for agentic workloads.
Available today · products + APIs
Stronger reasoning over Flash. Pairs frontier capability with deeper deliberation for harder tasks.
Coming next month
Any input modality → any output modality, starting with video generation. The first true cross-modal model.
Today in Gemini app, Flow, YT Shorts · APIs rolling out
Reimagines video search — jumps directly to relevant sections and surfaces the best matching segments inside a video.
Testing now · broad U.S. rollout summer 2026
Personalized AI agents running 24/7 in the background to find what you need, on your behalf and under your direction.
Summer 2026 · AI Pro/Ultra subscribers
Search builds custom dynamic layouts and interactive visuals per query — the result page changes shape to fit the question.
Summer 2026 · free for all
Persistent, custom dashboards for longer-running tasks you return to repeatedly — Search remembers the brief.
Coming months · Pro/Ultra U.S. first
Your personal AI agent inside the Gemini app — navigates your digital life and takes action under your direction.
Trusted testers this week · Beta for Ultra (U.S.) next week
A personalized morning digest that synthesizes inbox, calendar, and tasks into the most important things.
Launching
Voice "brain-dump" a document — Gemini structures and writes it for you. Editing capabilities follow.
Summer 2026 · subscribers
AI image creation/editing that treats every element as an object, not a flat image. Move, replace, edit anything.
Trusted testers now · broader rollout summer 2026 (Pro/Ultra)
Powerful voice composition for email and note creation — speak the draft, Gemini writes it.
Summer 2026
Standalone desktop app — a central home for orchestrating agents across all sorts of tasks.
Available today
Spin up specialized subagents with cross-platform terminal sandboxing, credential masking, hardened Git policies.
Available today
Programmatic control over the agent harness — deploy on your own infrastructure with custom integrations.
Available today
One Gemini API call provisions a fully sandboxed remote agent — no infrastructure to manage.
Available today · Gemini API
Vibe-code Android apps with native Kotlin, Workspace integrations, one-click Cloud Run + Firebase deploy.
Available today
Plans and reasons through complex creative tasks across Google's creative tools, under your control.
Rolling out today · everyone
Proposed open web standard letting browser agents access structured JavaScript and HTML tools — a Model Context Protocol for the open web.
Origin Trial · Chrome 149
Integrates DOM elements into canvas with WebGL/WebGPU — searchable, accessible, interactable rendered experiences.
Origin Trial
Agent-accessible quality audits, debugging surfaces, and real-time optimization verification — DevTools the agent can call.
Available
Expert-vetted skills for agents building performant, accessible web experiences across 100+ use cases.
Early Preview
A new UI space where you view live updates and task progress of agents like Spark working on your behalf.
Coming later 2026
Stable tool letting AI agents leverage Android Studio capabilities — SDK downloads, device testing, build pipelines.
Stable
Open-sourced tools helping LLMs execute Android best practices for complex workflows and migrations.
Open source
LLM leaderboard measuring AI performance on Android development tasks — Gemma 4 now included.
Available
Android Studio feature converting React Native, web, or iOS code to native Kotlin Android apps.
Preview
Gentle Monster sunglasses with spoken Gemini help in-ear. Hands-free interaction with the model wherever you go.
Launching fall 2026
Heads-up display glasses showing the information you need, when you need it — Gemini, visible.
Audio launches first · display timeline unspecified
Shows whether content's origin is AI or a camera, and whether it's been edited with generative tools.
Expanding to Search + Chrome
OpenAI, Kakao, and Eleven Labs now adopting Google's SynthID watermark — provenance across model providers.
Announced today
Connects Antigravity-style agentic platforms to 30+ major life-science databases and tools.
Available today · GitHub + Antigravity
Suite of AI experiments to accelerate scientific research, building on Deep Think and Deep Research.
Express interest on Google Labs