I/O opens amid fresh scrutiny of Gemini's competitiveness

Google opens its annual I/O developer conference on May 19 with a familiar promise and a narrower question: whether Gemini is keeping pace with the top consumer and developer AI products. The company has spent months pushing Gemini deeper into Search, Android and its broader software lineup, while also updating the standalone Gemini app. In the final days before I/O, a new “Extended” thinking option and more third-party app hooks began appearing for some users, according to 9to5Google and Google support pages. Google has set I/O for May 19-20 at Shoreline Amphitheatre in Mountain View, California, and online. ### Why is Gemini’s app rollout getting attention right before I/O? May 17 brought one of the clearest pre-I/O signs that Google is still tuning the Gemini product surface, not just the underlying model. 9to5Google reported that some users are seeing a new “Thinking level” menu in the Gemini app, with “Standard” and “Extended” options when selecting Fast or Gemini 3.1 Pro. (blog.google) The same report said Google is also preparing more third-party app integrations. 9to5Google listed current third-party apps including GitHub, OpenStax, Spotify and WhatsApp, and said support documents point to Canva, Instacart and OpenTable as additions that are coming but not yet broadly rolled out. Google’s help pages already describe connected-app features for GitHub, Spotify, WhatsApp and Instacart. (9to5google.com) ### What is Google officially saying Gemini can do now? Google DeepMind’s current Gemini pages describe Gemini 3 as the company’s “most intelligent AI model” and position Gemini 3.1 Pro for complex tasks, planning and building. The company says Gemini 3.1 Pro offers advanced multimodal understanding, stronger tool use and improved agentic capabilities for multi-step tasks. Google’s Gemini blog hub says the Gemini app has already received a major Gemini 3 update, and Google has separately promoted Gemini Agent and other app features in recent product posts. (9to5google.com) The company has also tied Gemini closely to I/O itself, saying the conference will feature updates “from Gemini to Android and more.” ### Where does the competitive pressure come from? (deepmind.google) April 23 gave Google another fresh comparison point when OpenAI introduced GPT-5.5 and said the model was built for coding, research, data analysis and work across tools. OpenAI said GPT-5.5 was rolling out to Plus, Pro, Business and Enterprise users in ChatGPT, and later to the API, while its pricing page lists GPT-5.5 at $5 per 1 million input tokens and $30 per 1 million output tokens. (blog.google) February 17 gave Anthropic its own update when it introduced Claude Sonnet 4.6 and said the model improved coding, computer use, long-context reasoning, agent planning and knowledge work. Anthropic says Sonnet 4.6 is the default model for some Claude users and offers a 1 million-token context window in beta. ### What, specifically, does Google need to show at I/O? Google has already put Gemini into multiple products, but the clearest questions at I/O are likely to be about performance, availability and execution. (openai.com) The company’s own Gemini pages emphasize reasoning, multimodality, tool use and agentic workflows, while the app changes reported this week point to more control over response depth and more actions across outside services. (anthropic.com) Support pages offer one concrete test of that product direction. Google documents connected-app use cases that include helping with code from GitHub, playing media from Spotify, sending messages through WhatsApp on Android and adding groceries to an Instacart cart. Those are the kinds of tasks Google can point to if it wants to argue Gemini is becoming more useful beyond chatbot answers. (deepmind.google) ### When will Google have to answer those questions in public? May 19 is the next fixed date. Google says I/O 2026 runs May 19-20 at Shoreline Amphitheatre in Mountain View, with keynote addresses, product demos and online sessions available through io.google. (blog.google) (support.google.com)