Google deprecates Gemini Omni Flash preview endpoint

gemini-omni-flash-preview is listed for Sept. 30, 2026; the GA successor is gemini-omni-1.1-flash

Google’s Gemini API changelog lists Wednesday as the deprecation date for gemini-omni-flash-preview. The Aug. 27, 2026, release note is short. “The existing gemini-omni-flash-preview endpoint will be deprecated on September 30, 2026.”

That sentence is the official mark. It sits under the same note that shipped the generally available successor, gemini-omni-1.1-flash.

Google’s deprecations page draws a line between the two words developers keep mixing up. “A ‘deprecation’ is the announcement that we no longer provide support for a model, and that it will be ‘shut down’ in the near future.” After shutdown, the company says, a model “is completely turned off, and the endpoint is no longer available.”

The Omni Flash table on that page lists gemini-omni-1.1-flash with an Aug. 27, 2026, release date and no shutdown date announced. The preview ID is the one the changelog put on the calendar.

Preview status was always temporary. Google’s model-version page says preview models “will be deprecated with at least 2 weeks notice.” The Aug. 27 note gave more than that.

The preview itself is three months old. On June 30, 2026, Google wrote: “Released gemini-omni-flash-preview, a high-performance multimodal model designed for high-speed video generation and conversational video editing.” The same entry said callers could use the Interactions API to “generate 3–10 second videos at 720p from text descriptions or animate still images, and then conversationally edit and refine the outputs.”

The GA note on Aug. 27 called gemini-omni-1.1-flash “the GA version of our fast, conversational video generation and editing model.” It added three capabilities the preview note did not list in that form.

Video extension: “Seamlessly extend existing videos by generating continuations at the end of a clip using the extend task or directly with a prompt.”

Interpolation: “Generate a video transitioning between two images using the image_to_video task with up to 2 images.”

Resolution control: “New resolution parameter in video_config supports 360p, 720p (default), 1080p, and 4k outputs. 1080p and 4K outputs are generated using upscaling.”

The current Gemini Omni Flash model page points production traffic at gemini-omni-1.1-flash. It still names gemini-omni-flash-preview as the preview ID. The page describes the GA model as “a high-performance model for fast, conversational video generation and editing.” Context window: 1,048,576 tokens. Output video: 3s-10s. Frame rate: 24 fps. Inputs: text, image, and video up to 10 seconds for editing and extension. Output: video.

The Omni guide adds length rules that matter once teams leave the preview string. “You can extend videos by 10s, up to a total length of 40s.” “Extension is limited to appending to the end of a video; prepending or extending the middle of a clip is not supported.” Tasks listed in video_config include text_to_video, image_to_video, reference_to_video, edit and extend.

Google’s own docs send developers to two pages to finish the swap: the Gemini Omni Flash model page and the omni guide. The Interactions API is the path both the June preview note and the current guide describe for conversational edits. A follow-up turn keeps the prior interaction id. The Files API is the documented route for uploading a clip the model did not generate.

The changelog does not publish a private-arrangement clause. It does not name a grace window past Sept. 30. It names the new ID.

For executives, the operational point is the model string in production configs, job templates and SDK defaults. For developers, the search is literal: gemini-omni-flash-preview. Replace it with gemini-omni-1.1-flash. Then test the resolution field and the extend task against the workload that used to assume a 720p, 3-to-10-second preview clip.

The Aug. 27 note is where Google said the preview endpoint would be deprecated. Wednesday is that date.