Google Flow appears to be testing Gemini Omni Flash and Agent Mode more widely, but the sharper story is not blanket access. It is Google's latest push to make Gemini feel like a working creative assistant.
Google I/O 2026 opens on May 19, and Flow is already giving the market something to watch. Users have reported seeing Gemini Omni Flash and Agent Mode inside Google's AI filmmaking tool, a sign that the company is pushing Gemini deeper into creative workflows before the keynote cycle fully plays out.
That matters because the original claim that Gemini Omni Flash is available in Flow for all users goes further than the evidence supports. Google's own Flow Help Centre describes Flow as an AI creative studio built around Veo, Imagen, and Gemini, but it also says access is tied to Google AI Pro and Ultra subscriptions. In other words, this looks current and relevant, but it should be treated as a wider Labs rollout or visible test, not a confirmed free release for everyone.
The bigger point still holds. Google is trying to make its AI products feel less like scattered model demos and more like tools that can carry a user through a job. Flow was introduced as a place to create, refine, and organize AI-generated video work. Adding an agent layer to that environment would make the product more useful, because the model would not only respond to prompts, but help plan scenes, adjust assets, and keep a project moving through multiple steps.
That is where the market is heading. Creative teams do not need another text box with better branding. They need systems that can preserve context, understand intent, and make revisions without forcing the user to rebuild the whole prompt every time. If Gemini Omni Flash can combine faster multimodal understanding with more reliable editing behavior, Flow becomes more than a showcase for Google's media models. It becomes a workspace.
Google's naming also points to the pressure it is under. Flash has usually signaled speed and lower-cost inference in the Gemini lineup, while Omni suggests broader multimodal capability. Put those ideas together and the pitch is clear: a model that can handle richer inputs without making every interaction feel expensive or slow. That is important for startups and enterprise teams, because agentic software only works at scale if the economics make sense in daily use.
There is still a gap between the product story and the verified facts. The available evidence does not confirm a formal Google announcement on May 15 that launched Gemini Omni Flash to all Flow users. It shows a current wave of user reports and a live Google Labs product that already combines Gemini with Veo and Imagen. That distinction matters. A publication can call the move significant without overstating what Google has officially shipped.
Why developers should care
For developers, the interesting part is not the model name. It is the interface pattern. If Flow adds a stronger agent mode, Google is signaling that the next competitive layer in AI will sit above the model itself. The useful product will be the one that can take initiative, understand the state of a project, and help the user make progress without turning every step into a fresh instruction.
That puts Google in a direct contest with OpenAI and Anthropic, but the fight is becoming more specific. OpenAI has pushed hard on broad multimodal capability, while Anthropic has built a strong position around reliable assistant behavior and enterprise trust. Google's advantage is distribution. If it can put capable agents inside products people already use, from Flow to Workspace to Android, it can make Gemini feel less like a separate destination and more like part of the operating layer.
As Reuters reported last year, Google has been working toward a more universal AI agent that can complete tasks on behalf of users. Flow gives that ambition a practical creative surface. A video tool is a natural place to test agent behavior because the work is iterative, visual, and full of small decisions. The model has to understand not only what the user asked for, but what changed, what still looks wrong, and what should happen next.
For startups, the takeaway is straightforward. Agentic AI is moving from chat windows into production tools, and that changes how products should be designed. The model is no longer just the engine under the hood. It can become part of the workflow itself, shaping how users move from idea to output.
That is why this moment is worth watching even before Google gives the full I/O picture. If Gemini Omni Flash and Agent Mode become a stable part of Flow, Google will have a clearer story around creative agents. The next question is whether it can bring the same coherence to the rest of Gemini, where developers and customers still need fewer names, clearer access, and products that prove their value in real work.
Also read: ICE moves to list futures on GPU compute, putting AI infrastructure on the trading floor • Google turns Gemini into its latest bet on a unified AI stack • Andrej Karpathy joins Anthropic as the AI race tightens