At its annual I/O 2026 developer conference, Google introduced new generative artificial intelligence models designed to advance AI agents and multimodal creation. The company unveiled Gemini 3.5 Flash, a fast reasoning model, and Gemini Omni, a creative model capable of generating and editing video.
Gemini 3.5 Flash is positioned as a more efficient coding and agentic AI model that can autonomously execute complex tasks and build software from scratch. This model powers Gemini Spark, an always-on, 24/7 personal AI assistant that proactively performs tasks for users to help them navigate their digital lives. Built on the Antigravity platform and harness, Gemini Spark integrates natively with Gmail and Workspace and runs on dedicated cloud VMs. It will be available to AI Ultra subscribers starting next week.
Additionally, Google DeepMind introduced the Omni family, featuring Gemini Omni. This multimodal model can generate and edit video using any combination of text, image, audio, and video inputs. While speech-editing capabilities are being withheld, the model includes SynthID watermarking by default.