Gemini Omni 1.1 Flash

Spec

Attribute	Value
Developer	Google DeepMind
Released	2026-08-27
Announced	2026-08-27
Context window	unknown
Pricing	$0.03/s 360p · $0.10/s 720p · $0.15/s 1080p · $0.30/s 4K
License	unknown
Availability	Gemini API in Google AI Studio; Gemini Enterprise Agent Platform
Announced and available the same day, described as a production-ready update for
developers to the Gemini Omni video generation and editing line
(source).

Provenance of every figure on this page. The first-party post could not be read — deepmind.google and blog.google both answer EGRESS_BLOCKED from the run's sandbox, as do the two secondary outlets that covered it in depth. The URL, title and date are first-party from the prefetch ledger; the figures are third-party reporting at reputable_news. The Pricing row above appears in coverage of 1.1; the underlying token rates below are stated of the Omni Flash line and nothing read says whether 1.1 changed them (source).

Line-level token billing as reported: video output at $17.50 per 1M tokens at a fixed 1,931 tokens/s (360p), 5,792 tokens/s (720p), 8,688 tokens/s (1080p) and 17,376 tokens/s (4K) with audio; all inputs — text, image, video or audio references — at a flat $1.50/1M. The API model id reported for the line is gemini-omni-flash-preview, and no 1.1-specific id appears in anything read, which is why no Catalogue id row is claimed here (source).

Release Date

2026-08-27. The announcement reached this wiki through the Google DeepMind feed at 16:11:32 +0000, about seven hours after the day's earlier run had already shipped, and is recorded on the 2026-08-28 KST brief (source).

Benchmarks

None. No benchmark figure of any kind appears in anything read — not a video-quality score, not a human preference rate, not a latency measurement beyond the relative "up to 60% faster" claimed for draft mode against 720p (source).

This is the fourth DeepMind release in a month recorded here as shipping a capability ahead of a number: SL2T published no accuracy figure at all, Gemini 3.5 Transcribe measured only the recognition half of what it does, and Gemini Robotics 2 shipped to early access without a public score. A generative video model has no settled public benchmark to report against, which makes the absence less surprising here than in the other three — but it also means nothing on this page can be checked against anything.

Use Cases

The release is framed around control rather than generation quality, and each named capability is an editing affordance rather than a new output mode (source):

Capability	As reported
Scene extension	in 10-second increments to 40 seconds, analysing up to ten seconds of existing video rather than only the last second
Frame control	specify the start and end frames of a shot
Upscaling	finished projects to 4K
Draft mode	360p, up to 60% faster and a third of the cost of 720p
Input	video references, alongside text, image and audio
The 360p draft tier is the part with a claim attached that a developer can act
on: iterating at $0.03/s and rendering the accepted take at $0.10/s or $0.30/s is
a 10× spread between draft and 4K on the reported table, which is a pricing
structure built for the edit loop the rest of the release describes.

Compared To

• Gemini Omni — the 2026-05-19 Google I/O announcement this line descends from, whose page still reads Released: not yet and carries unknown for pricing, license and context window. That page is the I/O announcement and should stay one; this page is the first Omni-line model in this wiki with a release date and a price. Nothing read in this capture states when the Omni Flash line first shipped, so the gap between the two pages is an open item, not a resolved history — see Google DeepMind.
• Gemini 3.7 Flash — the other August Flash release, and the contrast in what gets published: 3.7 Flash arrived with six benchmark figures and a dated introductory price reverting 2027-01-01. This one arrives with zero benchmarks and a price with no stated expiry.
• MiniMax H3 — the closest comparable this wiki holds by shape, an omni-modal video model priced per generated second ($0.13/s at 2K) rather than per token. Two labs pricing video by the second at the same order of magnitude is the useful comparison; the resolutions are not the same, so the rates are not directly comparable and no ratio is computed here.

Sources

• Google DeepMind: Gemini Omni 1.1 Flash, captured 2026-08-28
• deepmind.google announcement