I/O 2024 keynote
Sundar Pichai · Demis Hassabis · May 14, 2024 · 1h 53m · Mountain View, California
Summary
Google's 2024 keynote was about putting Gemini everywhere: AI Overviews in Search for everyone in the US, Gemini 1.5 Pro with a longer context in the Gemini app and for developers, the faster Gemini 1.5 Flash, and a first look at Project Astra, a universal AI agent. New media models (Imagen 3, Veo) and the Trillium TPU rounded it out.
Key themes: Models, Assistants, Agents · AI-written summary from the published transcript
- 1AI Overviews, the revamped Search results with generative answers, began rolling out to everyone in the US5:28
- 2Ask Photos, searching your photo library by asking Gemini, was set to roll out that summer7:37
- 3Gemini 1.5 Pro with a 1 million token context came to consumers in Gemini Advanced, with 2 million tokens for developers in private preview11:44
- 4Demis Hassabis showed Project Astra, a prototype universal AI agent that sees and remembers through the camera24:49
- 5Veo, a generative video model making 1080p clips, and Imagen 3 were introduced for creators34:04
- 6Trillium, the sixth-generation TPU, promised 4.7x the compute per chip of the previous generation39:07
- 7Gemini 1.5 Flash, a lighter and faster model, became available to developers in over 200 countries1:31:32
Announced
2 available the same day| Announcement | Status |
|---|---|
| AI Overviews in SearchRolling out to everyone in the US that week | GA |
| Gemini 1.5 Pro 2M context2 million tokens for developers, private preview | Preview |
| Project AstraUniversal AI agent prototype | Announced |
| Imagen 3Image generation model | Preview |
| Veo1080p video generation, select creators via VideoFX | Preview |
| Trillium TPUSixth-generation TPU, 4.7x compute per chip | Coming |
| Gemini 1.5 FlashFast, lower-cost Gemini model | GA |
| Gemma 2Next generation of the open Gemma models | Coming |
| SynthID for text and videoWatermarking extended to two new modalities | Announced |
Related keynotes
| Date | Keynote | |||
|---|---|---|---|---|
| May 19, 2026 | Google I/O 2026Sundar Pichai · 1h 51m · 8 announcedModels, Assistants, Agents | Models, Assistants, Agents | Gemini 3.5 Flash launched, combining frontier intelligence with agentic action for coding tasks | 8 |
| Apr 22, 2026 |  Google Cloud Next 2026Thomas Kurian, Sundar Pichai · 1h 39m · 8 announcedAgents, AI infrastructure, Security | Agents, AI infrastructure, Security | Thomas Kurian outlined a unified stack of chips, models, agents and security for enterprise AI | 8 |
| May 20, 2025 | Google I/O 2025Sundar Pichai · 1h 55m · 8 announcedModels, Assistants, Agents | Models, Assistants, Agents | Google now processes over 480 trillion tokens per month across its products | 8 |
| Apr 9, 2024 | Google Cloud Next 2024Thomas Kurian, Sundar Pichai · 1h 41m · 8 announcedModels, Agents, AI infrastructure | Models, Agents, AI infrastructure | NVIDIA Grace Blackwell (B200 and GB200) comes to Google Cloud early in 2025; TPU v5p became generally available | 8 |
About this keynote
Keynote page: www.youtube.com/watch?v=XEzRZ35urlk
- Sundar PichaiCEO of Google and Alphabet
- Demis HassabisCEO, Google DeepMind
Companies named: Google
Last changed 2026-09-24.
Sources: official keynote videos, their captions and the hosts' recaps, summarized by AI; watch the source for exact wording. Logos via logo.dev; trademarks belong to their owners.