{ "schemaVersion": "bitshovel-editorial-review-v1", "id": "reading-quality-91-20260912", "signalId": "alphasignal-alphasignal-ai-news-google-deepmind-s-gemini-3-7", "originalUrl": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-agentic-video-in-gemini/", "kind": "editorial-correction", "reviewedAt": "2026-09-12T13:30:20.540Z", "before": { "id": "alphasignal-alphasignal-ai-news-google-deepmind-s-gemini-3-7", "state": "new", "lane": "platformShift", "publishedAt": "2026-09-01T17:29:53.000Z", "observedAt": "2026-09-01T22:00:44.699Z", "originalUrl": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-agentic-video-in-gemini/", "relatedSources": [ { "label": { "zh": "AlphaSignal", "en": "AlphaSignal" }, "url": "https://alphasignal.ai/news/google-deepmind-s-gemini-3-7-flash-cuts-video-analysis-costs-by-66" }, { "label": { "zh": "Google DeepMind 博客", "en": "Google DeepMind blog" }, "url": "https://deepmind.google/blog/introducing-agentic-video-in-gemini/" } ], "source": { "zh": "alphasignal-daily 快照与原文", "en": "alphasignal-daily snapshot and original" }, "title": { "zh": "Gemini 3.7 Flash 上线 agentic 视频理解，成本最高降 66%", "en": "Gemini adds agentic video understanding at up to 66% less cost" }, "summary": { "zh": "Google 在 Gemini 3.7 Flash、3.6 Flash、3.5 Flash-Lite 上线 agentic 视频理解：模型用原生视频工具在视觉帧、音频与转写之间动态检索、扫描与检查目标片段，取代固定帧率（默认 1 FPS）的静态采样；Google 自报标准基准上分析成本最高降 66%、token 消耗最高降 88%、准确率最高升 7%，长视频（10 分钟教程到 90 分钟讲座）增益最明显。能力含秒内时刻检索、异常检测与精确计数，今日起经 Google AI Studio 与 Gemini Enterprise Agent Platform 用于视频上传与 YouTube。", "en": "Google launched agentic video understanding across Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite: the models use native video tools to dynamically search, scan, and inspect target segments across visual frames, audio, and transcripts, replacing fixed-rate (default 1 FPS) static ingestion. Google reports up to 66% lower analysis costs, up to 88% fewer tokens, and up to 7% better accuracy on standard benchmarks, with the largest gains on long-form video. Sub-second moment retrieval, anomaly detection, and precise counting, live in Google AI Studio and the Gemini Enterprise Agent Platform." }, "why": { "zh": "Gemini 主线 Flash 系首次以能力级事件进入视野：三款在役模型同批获得动态视频分析，把视频理解的单价曲线一次性拉低——对做视频批处理、监控与检索的团队是即时的成本结构变化。", "en": "The Gemini mainline Flash series enters the radar with a capability-level event: three in-service models gain dynamic video analysis at once, pulling down the unit-cost curve for video understanding — an immediate cost-structure change for teams doing batch video, monitoring, or retrieval." }, "caveat": { "zh": "证据来源为 Google 官方博客（厂商自述口径）与 AlphaSignal 同日聚合（BitShovel 于 2026-09-02 抓取核对原文）。“最高 66%/88%/7%”是 Google 自报基准上界，非独立复测；能力上手成本与真实任务表现仍待验证，热度不是实际适用性的证明。", "en": "Evidence is Google's official blog (vendor's own account) plus same-day AlphaSignal aggregation (original fetched and checked by BitShovel on Sep 2, 2026). The \"up to 66%/88%/7%\" figures are vendor-reported benchmark ceilings, not independent reruns; integration effort and real-task performance remain unverified, and attention is not proof of fit." }, "timeline": [ { "date": { "zh": "9 月 2 日", "en": "Sep 2" }, "datetime": "2026-09-01T22:00:44.699Z", "summary": { "zh": "进入 BitShovel 雷达：AlphaSignal 每日 收录时可见 100 条。", "en": "Entered the BitShovel radar: 100 entries visible on AlphaSignal daily at collection." }, "source": { "zh": "AlphaSignal 每日 快照", "en": "AlphaSignal daily snapshot" } }, { "date": { "zh": "9 月 2 日", "en": "Sep 2" }, "datetime": "2026-09-01T17:08:51.000Z", "summary": { "zh": "新增同题佐证：Google DeepMind 博客 于本观察窗收录同一故事（标题匹配）。", "en": "Corroboration added: Google DeepMind blog carried the same story this window (title match)." }, "source": { "zh": "Google DeepMind 博客", "en": "Google DeepMind blog" }, "kind": "corroboration-added", "discoveredAt": "2026-09-04T04:00:26.282Z" } ], "followTarget": { "id": "discussion:blog-google-agentic-video-in-gemini", "label": { "zh": "Gemini agentic 视频理解官方公告", "en": "Official Gemini agentic video announcement" }, "scope": { "zh": "agentic 视频能力的后续定价、限额与模型覆盖修订，以及 Flash 系列新版本动态", "en": "Follow-up pricing, quota, and model-coverage revisions to agentic video, plus new Flash-series releases" }, "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-agentic-video-in-gemini/", "updateMode": "verified-change" }, "ingestion": { "mode": "editorial", "adapter": "syndication-feed", "evidenceClass": "official-statement", "observationRef": "source-observations/history/syndication-feeds.2026-09-01T22-00-44Z.snapshot.json#feed:alphasignal-daily", "changeRef": "source-observations/history/syndication-feed-changes.2026-09-01T22-00-44Z.snapshot.json#feed:alphasignal-daily" }, "managedObjectId": "model-version:gemini-3-7-flash", "banner": { "url": "/card-art/gemini-agentic-video-official.webp", "credit": "Google", "license": "© Google - editorial use with attribution" }, "verifiedAt": "2026-09-08T16:00:21.541Z", "retireReceipt": { "decision": "retire-candidate", "basis": "silent", "lateral7d": 0, "vertical7d": 0, "decidedAt": "2026-09-09T06:24:49.890Z" }, "__retiredAt": "2026-09-09T06:24:49.890Z" }, "after": { "title": { "zh": "Gemini 视频理解：按问题找片段，费用省在少读材料", "en": "Gemini video understanding: find the relevant moments, process less" }, "summary": { "zh": "Gemini 可以根据问题选择视频片段、画面、音频和转写，再继续检查细节。长视频分析因而有机会减少处理量；官方所说的降本来自使用方式变化，标准 token 单价没有因此下降。", "en": "Gemini can choose video segments, frames, audio and transcripts in response to a question. Processing less material can reduce long-video costs; standard token prices have not been cut by this feature." }, "why": { "zh": "围绕问题读取相关片段，可以减少长视频处理量，适合进一步比较质量、费用和等待时间。", "en": "Question-directed segment loading can reduce long-video processing; quality, cost and latency still need comparison." }, "caveat": { "zh": "降本与准确率增益为官方基准上限，不是本站实测；标准 token 单价未因该功能下降，产品接入计划与已可用接口分开表述。", "en": "Cost and accuracy improvements are vendor benchmark maxima, not BitShovel measurements. Standard token rates are unchanged; future product integrations are separate from available APIs." } }, "note": { "zh": "分开处理费用与 token 单价，补充视频模式选择、动态读取与当前文档支持范围。 保留事件、发现和收录日期。", "en": "Separated processing cost from token rates and added mode choice, dynamic loading and current documentation scope. Event, discovery and collection dates remain unchanged." }, "withdrawnSources": [], "withdrawnTimeline": [], "evidence": [ { "type": "retained-official-page", "label": { "zh": "alphasignal-daily 快照与原文", "en": "alphasignal-daily snapshot and original" }, "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-agentic-video-in-gemini/", "fetchedAt": "2026-09-12T13:16:16.001Z", "path": "app/_content/editorial-reviews/evidence/reading-quality-91-95-5a222bc71ad3-20260912.json", "bytes": 661, "sha256": "b1814ca0ef812df98122ed8746b22aa528670cd4e19e854c48fe27c4155460b6" }, { "type": "retained-official-page", "label": { "zh": "Gemini 视频理解文档（英文）", "en": "Gemini video understanding documentation (English)" }, "url": "https://ai.google.dev/gemini-api/docs/video-understanding?hl=en", "fetchedAt": "2026-09-12T13:25:52.062Z", "path": "app/_content/editorial-reviews/evidence/reading-quality-91-95-b532d62311c0-20260912.json", "bytes": 571, "sha256": "92db5c1acc8d5d5a72703d999668098485ef5725aaa0543f3eed55cf4527e56f" } ], "integrity": "b5c0b65d28972276841fee0178527a42aebecb6c9867db126ff8e67faa46be02" }