{"id":116011,"date":"2026-09-06T01:24:20","date_gmt":"2026-09-06T01:24:20","guid":{"rendered":"https:\/\/youzum.net\/google-launches-agentic-video-understanding-for-gemini-flash-models-cutting-video-tokens-by-up-to-88\/"},"modified":"2026-09-06T01:24:20","modified_gmt":"2026-09-06T01:24:20","slug":"google-launches-agentic-video-understanding-for-gemini-flash-models-cutting-video-tokens-by-up-to-88","status":"publish","type":"post","link":"https:\/\/youzum.net\/ja\/google-launches-agentic-video-understanding-for-gemini-flash-models-cutting-video-tokens-by-up-to-88\/","title":{"rendered":"Google Launches Agentic Video Understanding for Gemini Flash Models, Cutting Video Tokens by Up to 88%"},"content":{"rendered":"

Video has been the most expensive modality to reason over. A Gemini model handed a 90-minute lecture has, until now, ingested the whole thing at a fixed one frame per second, whether the question was \u2018summarize this\u2019 or \u2018what time does the speaker switch to the pricing slide?\u2019 That single-pass design forces a bad trade: pay for the full timeline in context, or pre-chunk the video and risk dropping the detail that mattered.<\/p>\n

This week, Google launched agentic video understanding<\/a> across its Flash models. Instead of ingesting the timeline, Gemini navigates it deciding what to watch, at what frame rate, and through which modality. Google reports up to 88% fewer tokens, up to 66% lower cost, and up to 7% higher accuracy on standard video benchmarks.<\/p>\n

Is it deployable?<\/strong> Yes, but only as a hosted API feature. There are no open weights and nothing to self-host. It ships through the Gemini API<\/a> in Google AI Studio and the Gemini Enterprise Agent Platform<\/a>, works with both file uploads and public YouTube URLs, and bills at standard Gemini API token pricing with no additional feature fee.<\/p>\n

What actually changed<\/strong><\/h2>\n

Static processing, still the default on every Gemini model, extracts frames at 1 FPS in a single pass, processes audio at 1 Kbps single channel, and inserts timestamps every second. Agentic processing replaces that with a loop. The model pairs its own reasoning with native video tools to search, scan, and inspect target segments across frames, audio, and transcripts, loading only what the prompt requires. Developers could already assemble this by hand; the change is that Gemini runs the loop internally, which is where the development overhead disappears.<\/p>\n

Across Google\u2019s evaluations, Gemini 3.7 Flash with agentic understanding lands on the accuracy-to-cost Pareto frontier for video analysis among the models tested. The efficiency gains concentrate on long-form content, from 10-minute how-to guides to multi-hour recordings.<\/p>\n

\n<\/div>\n

\n

What the API returns<\/strong><\/h2>\n<\/p>

Agentic processing adds two step types to the response steps<\/code> array: a processing_call<\/code> when the model requests a segment or transcript, and a matching processing_result<\/code> when that load completes. They interleave with thought<\/code> steps and precede model_output<\/code>, so they can drive a live progress trace in your UI. Their presence is also how you verify agentic mode actually ran.<\/p>\n

Token accounting splits accordingly. Navigation reasoning bills as thought tokens (total_thought_tokens<\/code>); frames, audio, and transcripts loaded on demand bill as tool-use tokens (total_tool_use_tokens<\/code>).<\/p>\n

Enabling it is one field on the video part:<\/p>\n

\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
interaction = client.interactions.create(\n    model=\"gemini-3.7-flash\",\n    input=[\n        {\n            \"type\": \"video\",\n            \"uri\": \"https:\/\/youtu.be\/7Z5Vy9JBANs\",\n            \"processing\": \"agentic\"\n        },\n        {\n            \"type\": \"text\",\n            \"text\": \"What are the 3 most important announcements in this keynote?\",\n        },\n    ],\n)<\/code><\/pre>\n<\/div>\n<\/div>\n
You can also mix modes per video inside a single request, agentic on the long lecture, static on the short clip.<\/p>\n
Key Takeaways<\/strong><\/h2>\n