◉GEMINI LABJP
●OMNI — Gemini Omni Flash, a natively multimodal model, enters API public preview for building custom video workflows●NANO — Nano Banana 2 Lite arrives as the fastest and most cost-efficient Gemini image model yet●FLASH — Gemini 3.5 Flash reaches general availability with sustained frontier performance on agentic and coding tasks●AGENTS — Managed Agents enter public preview in the Gemini API, running autonomous, stateful agents in isolated Google-hosted Linux sandboxes●MEMORY — The Memory Bank IngestEvents API is GA, decoupling event ingestion from memory generation for continuous streaming●THROUGHPUT — Provisioned throughput now accepts up to seven pending model orders for the same model and region●OMNI — Gemini Omni Flash, a natively multimodal model, enters API public preview for building custom video workflows●NANO — Nano Banana 2 Lite arrives as the fastest and most cost-efficient Gemini image model yet●FLASH — Gemini 3.5 Flash reaches general availability with sustained frontier performance on agentic and coding tasks●AGENTS — Managed Agents enter public preview in the Gemini API, running autonomous, stateful agents in isolated Google-hosted Linux sandboxes●MEMORY — The Memory Bank IngestEvents API is GA, decoupling event ingestion from memory generation for continuous streaming●THROUGHPUT — Provisioned throughput now accepts up to seven pending model orders for the same model and region
TAG

thinking-level

1 articles
← Back to all tags
Related:
gemini1latency1cost1mobile1
⬡ Gemini Advanced/2026-07-12Advanced

Spend Deep Reasoning Only Where It's Needed: Per-Request thinking_level Routing in Gemini

Running every request at high thinking_level bloats latency and cost; forcing low drops accuracy on hard questions. This walks through a router that picks Gemini 3.x thinking_level per request from an inexpensive difficulty estimate, keeping p95 latency inside a mobile budget while reserving deep reasoning for the questions that need it — with measured numbers and working code.