GEMINI LABJP
MCPCONN — Gemini in Google Workspace now connects to Asana, Atlassian Rovo, HubSpot, Mailchimp, QuickBooks, Monday and Salesforce over MCP. It is on by default and managed from the admin consoleGLOBAL — The Gmail Search AI Overviews rollout described as global comes with conditions. The announcement itself requires English as the display language and excludes personal accounts in Japan10/02 — gemini-2.5-flash-image shuts down on October 2, ten days away. The replacement is gemini-3.1-flash-image-preview402 — Depleted prepay credits now return HTTP 402 instead of 429. The status field still reads RESOURCE_EXHAUSTED, so retry logic that branches on status will never stopNEW — The day I first disconnected an MCP server: how I choose which tools stay enabledSKILLS — In-app notices point from Gems to Skills, but Skills is described as unavailable on work and school accounts, and the dates have not been confirmed officiallyMCPCONN — Gemini in Google Workspace now connects to Asana, Atlassian Rovo, HubSpot, Mailchimp, QuickBooks, Monday and Salesforce over MCP. It is on by default and managed from the admin consoleGLOBAL — The Gmail Search AI Overviews rollout described as global comes with conditions. The announcement itself requires English as the display language and excludes personal accounts in Japan10/02 — gemini-2.5-flash-image shuts down on October 2, ten days away. The replacement is gemini-3.1-flash-image-preview402 — Depleted prepay credits now return HTTP 402 instead of 429. The status field still reads RESOURCE_EXHAUSTED, so retry logic that branches on status will never stopNEW — The day I first disconnected an MCP server: how I choose which tools stay enabledSKILLS — In-app notices point from Gems to Skills, but Skills is described as unavailable on work and school accounts, and the dates have not been confirmed officially
TAG

p95

1 articles
Back to all tags
Related:
Gemini API1Tail Latency1SLO1Streaming1
Gemini API/2026-06-23Advanced

Your Gemini API Average Latency Looks Great — But Some Users Still Get Stuck. Defending p95/p99

Your average TTFT is fast, yet a fraction of users keep hitting frozen responses. That is a tail-latency problem (p95/p99). From measurement to model routing, streaming budgets, cache accounting, and retry design — here are the defenses that actually held up in production, with code.