Google used its I/O 2026 keynote on 19 May 2026 to introduce a new generation of Gemini models and a wave of agent-focused products. Chief executive Sundar Pichai framed the announcements as the start of what he called an "agentic era".
The new Gemini models
The headline release was Gemini 3.5 Flash, Google's latest Flash-tier model and the first generally available entry in the Gemini 3.5 family. Google said it outperforms the previous Gemini 3.1 Pro on coding and agentic benchmarks, and runs roughly four times faster than other frontier models in output tokens per second. Liz Reid, Google's head of Search, described it as "delivering sustained frontier performance for agents and coding". The model is available through the Gemini API, Google AI Studio, Android Studio, Google Antigravity, Gemini Enterprise, the Gemini app and AI Mode in Search.
A larger sibling, Gemini 3.5 Pro, is in internal testing. Google said the model is scheduled to roll out in June 2026 but did not give a specific date.
The second major release is Gemini Omni, a family Google described as combining "an intuitive understanding of physics" with the model's existing world knowledge. The first model in the family, Gemini Omni Flash, generates outputs in any modality from any input, starting with video. Each clip is limited to 10 seconds — TechCrunch reported that the cap is a deliberate product decision rather than a model constraint. Output includes a SynthID digital watermark. Omni Flash is rolling out to Google AI Plus, Pro and Ultra subscribers globally, with free access through YouTube Shorts' Remix feature.
The product wave around the models
Alongside the models, Google introduced several agent-focused products:
- Gemini Spark — a 24-hour personal AI agent that Google says runs on dedicated cloud virtual machines
- Ask YouTube — a conversational interface that lets users query video content and jump to specific timestamps
- Google Antigravity 2.0 — an "agent-first" development platform with a new CLI and SDK
- Universal Cart — a shopping hub that pulls inventory across Search, Gemini, YouTube and Gmail into a single checkout
- Daily Brief — a personalised digest agent
- Android XR — Google's smart eyewear platform, scheduled for an autumn 2026 launch
A new $100-per-month AI Ultra subscription tier was also introduced alongside the existing AI Pro and AI Plus plans. Google did not break down which features are gated to each tier.
The numbers Pichai cited
Pichai used the keynote to walk through several growth metrics:
- 3.2 quadrillion tokens processed monthly across Google's AI products — a seven-fold year-on-year increase
- 900 million monthly active users on the Gemini app — roughly double the figure from a year ago
- 8.5 million developers building with Google models each month
- 2.5 billion monthly active users for AI Overviews in Search
- 50 billion images generated with Google's Nano Banana family of models
- $180–190 billion in projected annual capital expenditure for 2026, up from $31 billion in 2022
Pichai said: "We're now in the part of the AI cycle where people want to see the value in the products they use every day."
What remains unclear
Google did not publish detailed methodology for the benchmark claims around Gemini 3.5 Flash. The exact release date for Gemini 3.5 Pro is unconfirmed beyond "June 2026", and the company has not disclosed whether Pro will use the same Omni multimodal stack. Pricing details for the new AI Ultra tier, including any usage caps, were not fully laid out in the keynote.
The longer Omni timeline is also open. Google said the family is "starting with video" and intends to add image and text generation later, but did not commit to a launch window for those modalities.
Most of the I/O 2026 announcements are scheduled to roll out through summer 2026. The first checkpoint is Gemini 3.5 Pro's general release, which Google has placed in June. Whether the headline benchmark gains hold up under independent third-party testing is the next thing worth watching.
