Breaking

Google's Gemini Omni Flash Takes Top Spot on Video Arena

July 2, 2026 at 15:08 EDT

Media Generation
Foundation Models

Google DeepMind's video generation model "Gemini Omni Flash" (model ID: gemini-omni-flash-preview) has claimed the overall No. 1 position on Video Arena, the human-voted ranking for video generation and editing. Its Elo score of 1404 puts it 101 points ahead of second-place Seedance 2.0 Mini from BytePlus.

June 30, 2026 · Google DeepMind

Gemini Omni Flash Seizes the No.1 Spot in Video Generation

Google's native-multimodal video model tops the human-preference Video Arena with an Elo of 1404 — leading the runner-up by 101 points in one of the platform's largest single-model leaps, and turning enterprise video production into a conversation.

1404

Overall Elo rating — No.1 on Video Arena

+101

Points ahead of 2nd place — a record-size leap

$0.10

Per second of output — clips up to ~10s with audio

Overall Video Arena — top of the leaderboard

Elo rating (chess-style, from blind human preference votes)

1404

Gemini Omni Flash

1303

Seedance 2.0 Mini

Category breakdown — leader vs Omni Flash

Text-to-Video · 1st

Gemini Omni Flash leads the field

1527

Image-to-Video · 2nd

Behind Seedance 2.0 720p (1474)

1469

Video Edit · 2nd

Behind Seedance (1377)

1347

Why it feels different

A native multimodal design — text, image, video and audio together — turns one-shot generation into a conversation.

Generate rough footage → Refine via dialogue → Swap · restyle · relight

Where it wins

Conversational, multi-turn editing
Editing speed & consistency across turns
Multimodal grounding & workflow integration

Where it lags

Seedance 2.0 stronger on one-shot cinematic realism
Complex motion, physics & long-edit consistency
Accurate text rendering still imperfect

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.