Google is adding agent-based video analysis to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite models. Rather than processing video frames at a fixed rate, the models now autonomously decide which segments to examine and at what resolution, resulting in (The Decoder) up to 88 percent reduction in token usage while maintaining or improving output quality. The capability addresses a structural efficiency challenge in video understanding workloads.
Anthropaic released Claude Fable 5.1 and Claude Mythos 5.1, with Fable 5.1 scoring (MarkTechPost) 52.6 percent on Terminal-Bench-Science and improving agentic coding by over 30 percent. Inference costs drop (The Decoder) by up to 45 percent for long autonomous runs with many tool calls, with cache read costs dropping 75 percent. Fable 5.1 is available on the Claude API, AWS, Google Cloud, and Microsoft Foundry. Separately, Perplexity shipped hybrid compute on Mac, allowing (MarkTechPost) cloud agents to orchestrate tasks down to local models for sensitive context, addressing privacy constraints around confidential documents and privileged files.