Releases·Model Release
Gemini's 88% video token cut landed on 3.7 Flash, not 3.8
Google shipped agentic video understanding to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite on September 1, one day before the Gemini 3.8 Flash release. Instead of fixed-rate frame sampling, the model chooses which segments to inspect and whether to read frames, audio, or the transcript. Google claims up to 88% fewer tokens and up to 66% lower cost at the same API pricing, a change a Reddit poster argues matters more for cost than 3.8's throughput gains.
2 engagement1 sourceTue, Sep 1, 2026, 12:00 AM
AI Overview
- Google shipped agentic video understanding to Gemini 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite on September 1, releasing 3.8 Flash the next day
- The model picks which segments to view and whether to read frames, audio or transcript instead of sampling every frame at a fixed rate
- Google claims up to 88% fewer tokens and up to 66% lower cost at the same API pricing
- Google also claims up to 7% higher accuracy, without specifying benchmarks; a Reddit commenter questions skipping models' recall on long footage
Related
Social·Community
User retrospective compares Gemini 3.5, 3.6, and 3.7 Flash ahead of expected 3.8 Flash
A heavy Gemini user posted a hands-on retrospective of the last three Flash generations on Reddit, anticipating Gemini 3.8 Flash. The comparison finds 3.5 Flash strongest at web search and grounding, while 3.7 Flash leads on writing and creative tasks. The thread offers practitioners subjective but daily-tested impressions of how each generation shifted capabilities.
A heavy Gemini user posted a hands-on retrospective of the last three Flash generations on Reddit, anticipating Gemini 3.8 Flash. The comparison finds 3.5 Flash strongest at web search and grounding, while 3.7 Flash leads on writing and creative tasks. The thread offers practitioners subjective but daily-tested impressions of how each generation shifted capabilities.
51 engagement1 sourceredditMon, Aug 31, 2026, 06:16 PM
Social·Community
Reddit users report Google removing Gemini models: 3.7 Flash gone after 3.8 launch, picker locked to 3.5 Flash-Lite
Reddit users across multiple posts report that Google removed selectable Gemini models — 3.5 Flash-Lite, 3.6 Flash, 3.1 Pro, and 3.7 Flash — replacing model choice with an 'Auto' mode plus Low/Medium/High labels that one user claims always routes to 3.5 Flash-Lite. A startup developer says `gemini-3.7-flash` API requests were silently rerouted to `gemini-3.8-flash`, degrading their form-field detection benchmark from 98–99% to 94–95%. Newer posts describe testing Google's Argon model on Pro plans with tight usage limits, while praising Claude Opus 5.5 by comparison.
Reddit users across multiple posts report that Google removed selectable Gemini models — 3.5 Flash-Lite, 3.6 Flash, 3.1 Pro, and 3.7 Flash — replacing model choice with an 'Auto' mode plus Low/Medium/High labels that one user claims always routes to 3.5 Flash-Lite. A startup developer says `gemini-3.7-flash` API requests were silently rerouted to `gemini-3.8-flash`, degrading their form-field detection benchmark from 98–99% to 94–95%. Newer posts describe testing Google's Argon model on Pro plans with tight usage limits, while praising Claude Opus 5.5 by comparison.
Updated Thu, Oct 15, 2026, 04:05 PM — Users report testing Argon on Pro plans: tight usage limits, unavailable in Antigravity, versus Opus 5.5 completing task
01561 engagement6 sourcesredditSat, Oct 10, 2026, 03:46 PM
Social·Community
Gemini 3.5 Flash-Lite suffices for most users, says developer of three-model deployment
A practitioner running a multi-model AI service for a large institution reports that most users choose Gemini 3.5 Flash-Lite over Gemini 3.8 Flash and Claude Sonnet 5 and express full satisfaction. The post argues entry-level models now exceed former frontier models like GPT-4, making larger models a niche necessity rather than the default for real-world workflows including legal work.
A practitioner running a multi-model AI service for a large institution reports that most users choose Gemini 3.5 Flash-Lite over Gemini 3.8 Flash and Claude Sonnet 5 and express full satisfaction. The post argues entry-level models now exceed former frontier models like GPT-4, making larger models a niche necessity rather than the default for real-world workflows including legal work.
16 engagement1 sourceredditSat, Oct 3, 2026, 04:21 PM
Developments·Benchmark
Community benchmarks of Gemini 3.6 Flash emerge
Reddit users are sharing and discussing benchmark results for Google's Gemini 3.6 Flash model. The posts indicate the model is now available for testing, with early performance numbers being reported.
Reddit users are sharing and discussing benchmark results for Google's Gemini 3.6 Flash model. The posts indicate the model is now available for testing, with early performance numbers being reported.
012293 engagement4 sourcesredditTue, Jul 21, 2026, 03:14 PM
Releases·Model Releasefeatured
Google launches Gemini 3.7 Flash for coding and agents at half the price
Google released Gemini 3.7 Flash, its latest workhorse model for coding and agent workflows, just three weeks after 3.6 Flash. It offers substantial improvements in software engineering and knowledge work at an introductory price half that of 3.6 Flash. The model is available in Google AI Studio, Vertex AI, and Gemini, with integrations like GitHub Copilot already supporting it.
Google released Gemini 3.7 Flash, its latest workhorse model for coding and agent workflows, just three weeks after 3.6 Flash. It offers substantial improvements in software engineering and knowledge work at an introductory price half that of 3.6 Flash. The model is available in Google AI Studio, Vertex AI, and Gemini, with integrations like GitHub Copilot already supporting it.
71539.4k engagement28 sourcesreddit, rss, deepmind.google +3Thu, Aug 13, 2026, 12:00 AM