Google's Gemini 3.5 Flash runs 4x faster, Omni generates physics-aware video
New model goes global today; Omni crafts video with simulated gravity and motion.
Google kicked off I/O 2026 with two major AI releases: Gemini 3.5 Flash and Gemini Omni. Gemini 3.5 Flash is now available globally, optimized specifically for agentic workflows and coding tasks. Google claims it runs four times faster than comparable frontier models, making it a strong contender for real-time applications, tool use, and multi-step reasoning. The model targets developers and enterprises seeking lower latency and higher throughput without sacrificing accuracy.
Gemini Omni, aimed at paid AI subscribers (Google One AI Premium), marks a leap into generative video. It fuses Gemini’s multimodal reasoning with generative media models—Nano Banana, Veo, and Genie—to accept any input (text, image, audio, video) and produce coherent video output. Notably, Omni simulates physical properties like gravity, kinetic motion, and object interactions, enabling realistic content without manual editing. This positions Omni as a creative accelerator for marketers, game developers, and filmmakers.
- Gemini 3.5 Flash delivers 4x faster inference than rival frontier models, ideal for agentic and coding workflows.
- Gemini Omni generates video from any input, simulating physics (gravity, motion) using a blend of Gemini and generative media models.
- Gemini 3.5 Flash is available globally today; Omni is live for Google One AI Premium subscribers.
Why It Matters
Google offers both speed for developers and realistic video generation, expanding AI’s practical use in enterprise and creative fields.