Google held I/O 2026 this week and shipped three distinct models: Gemini 3.5 Flash for agentic tasks, Gemini Omni for video generation, and an updated Antigravity platform that orchestrated 93 agents in a live demo. Not announcements about future capabilities - shipping APIs, available now.
Gemini 3.5 Flash: The Coding Model
Flash is positioned as the agentic model in Google's lineup. Faster inference, optimized for tool use and code generation, designed for the workflow where a model needs to chain multiple actions together. Think: "analyze this codebase, identify the bug, propose a fix, test it, commit if it passes".
What's interesting here is the positioning. Google isn't trying to make one model do everything. Flash is explicitly for agents and developers. The benchmarks focus on coding tasks, tool-use accuracy, and multi-step reasoning. They're optimizing for a specific use case rather than chasing general-purpose supremacy.
For developers building on Gemini, this gives you a clear choice: if you're building an agent that needs to interact with APIs and write code, you use Flash. If you need general knowledge or multimodal understanding, you use the standard Gemini model. Specialization over generalization.
Gemini Omni: Video Generation Goes Multimodal
Omni handles video generation, but not in the Sora "text-to-cinematic-video" sense. This is multimodal input to video output - feed it text, images, audio, existing video clips, and it generates new video content that combines those elements.
The demos showed video editing driven by natural language prompts, style transfer between clips, and object insertion that respects physics and lighting. Practical video editing tasks, not art projects.
The technical detail that matters: Omni uses what they're calling "NanoBanana" diffusion architecture. Smaller, faster, designed to run video generation at reasonable cost. We'll see if that claim holds when people start hitting the API at scale, but the intent is clear - make video generation cheap enough to use in production workflows.
Antigravity 2.0: The 93-Agent Demo
Antigravity is Google's agent orchestration platform. Version 2.0 launched with a demo that coordinated 93 agents simultaneously - different models, different specializations, all working on a complex multi-step task.
This isn't about one smart agent. It's about coordinating many specialized agents. One agent handles data retrieval, another analyzes it, another generates a report, another fact-checks the report, another optimizes the presentation format. Antigravity manages the workflow, handles failures, routes tasks to the right agent, and aggregates results.
The 93-agent demo was deliberately ambitious - pushing the platform to show what's possible when you stop thinking about "an AI assistant" and start thinking about "a workforce of specialized AI tools". Whether anyone needs 93 agents is debatable. Whether the orchestration layer can handle that complexity reliably is the real question.
What Developers Get
All three launches are immediately available through Google Cloud APIs. Flash pricing is positioned below GPT-4 for coding tasks. Omni video generation costs less than Runway or Pika for equivalent output. Antigravity orchestration is billed per agent-action, not per request.
For builders, this is infrastructure. Flash gives you a better model for agent workflows. Omni makes video generation economically viable for more use cases. Antigravity provides the plumbing to coordinate multiple models without writing your own orchestration layer.
The pattern across all three: Google is building for developers who want to ship products, not researchers experimenting with capabilities. These are tools designed for production use, priced to be adopted, and documented with the assumption you're building something real.