Google I/O 2026 opens Tuesday May 12 at the Shoreline Amphitheatre in Mountain View with the keynote at 10 a.m. pacific. The conference runs through Thursday May 14 with three days of developer sessions, Android workshops, and the public unveiling of what Google has been building since the Gemini 2 release in December. Sundar Pichai will deliver the keynote with Demis Hassabis of DeepMind and Rick Osterloh of devices and services. The full developer keynote follows on Wednesday May 13 at 9:30 a.m. and will cover Android 17, Chrome OS Flex 2, and the developer tooling around Gemini.
The biggest expected announcement is Gemini 3. Google has been seeding the model to select developers since early April under a non disclosure agreement and the early benchmark numbers are competitive with Claude Opus 4.6 and OpenAI's o4 series. The model reportedly improves on Gemini 2.5 in agentic workflow tasks, with SWE-Bench Verified scores in the 84 to 87 percent range and TAU-Bench scores above 76 percent. The pricing structure is expected to launch at parity with Anthropic's Sonnet 4.6 tier, which would put Google in a position to undercut OpenAI on enterprise cost while matching capability. Whether the model genuinely matches Claude Opus 4.6 across the full reasoning benchmark suite is the question developers will be testing against in the days after release.
The Pixel 11 and Pixel 11 Pro hardware announcements are expected to anchor the Wednesday session. Leaked images from a Vietnam factory floor in early April showed a redesigned camera bar that wraps the back of the phone and a new under display fingerprint sensor that uses ultrasonic technology rather than optical. The Pixel 11 Pro will reportedly ship with the Tensor G6 chip, which is the first Tensor chip to be manufactured on TSMC's N3 process rather than Samsung's. The shift is significant for performance and battery life, with leaked benchmarks showing roughly 28 percent improvement in single core performance and 41 percent improvement in sustained GPU performance. Pricing is expected to hold at 799 dollars for the standard model and 999 dollars for the Pro.
The deeper story at I/O 2026 is what Google calls Gemini OS, which is the company's positioning for Gemini as the layer underneath Android, Chrome, Workspace, and Search. The pitch from Google is that Gemini is no longer a feature inside other products but a fabric that runs across them. Practical examples include a redesigned Google Search experience where Gemini handles every query as a conversation, a Workspace integration that allows Gemini to draft, edit, and send emails on your behalf with explicit user confirmation, and an Android system layer that allows third party apps to call Gemini for free as part of the OS. The competitive frame is clear. Apple is building Apple Intelligence into iOS. Microsoft is building Copilot into Windows. Google is building Gemini into Android.
The developer story matters as much as the consumer story. Google AI Studio, the developer environment for Gemini, is expected to receive a significant overhaul that includes integrated agent tooling, the ability to deploy Gemini powered agents directly to Google Cloud Run, and a new pricing tier called Gemini Code that targets developers building with the model in IDE environments like Cursor and Claude Code. The Code tier would include unlimited access to Gemini 3 for code completion and 50 dollar per month at a flat rate, which is a more aggressive position than what Cursor charges for Claude or what GitHub Copilot charges for OpenAI models.
The Android 17 announcement is also expected to introduce significant changes to the system level integration of generative video and image features. The on device Gemini Nano 2 model is reportedly capable of generating short video clips at up to 480p directly on the Pixel 11 Pro without server connectivity, which would be the first major mobile platform to ship that capability. Apple has been rumored to ship a similar feature with iOS 19 at WWDC in June, and the timing competition is shaping the broader narrative around on device AI. Whether developers can call these models directly through public APIs will be the question.
The competitive landscape going into I/O is dense. Microsoft Build runs May 19 through 22, NVIDIA Computex runs May 19 through 23, and Apple WWDC follows June 8 through 12. The four week stretch from May 12 through June 12 will include almost every major AI and platform announcement of 2026, and Google is leading the calendar. The pressure on Google is to demonstrate that Gemini is genuinely competitive with Claude and OpenAI rather than just close enough. The early developer signal suggests they have closed the gap. The wider market signal will be tested over the four weeks that follow. For developers, the developer keynote Wednesday May 13 is the more actionable session.