Google has added Gemini 3.7 Flash as a selectable model in AI Mode, one day after the model\u2019s initial release for coding and agent tasks.<\/p>\n
Robby Stein<\/a>, VP of Product for Google Search, announced the rollout on X, writing that the model is \u201cbetter at following instructions + understanding your intent, so you get even more helpful responses.\u201d<\/p>\n You can choose Gemini 3.7 Flash from the model menu in AI Mode\u2019s \u201cAsk anything\u201d bar, which you can find by clicking the \u201c+\u201d icon. It\u2019s located under the \u201cGemini 3 models\u201d section, right next to \u201cAuto\u201d and \u201cPro.\u201d<\/p>\n Stein notes that the rollout is global and available today to Google AI Pro and Ultra subscribers in English. He didn\u2019t say whether Gemini 3.7 Flash becomes AI Mode\u2019s default model or whether it factors into the automatic model routing Google began building into AI Mode<\/a> last November.<\/p>\n Background<\/p>\n Google introduced Gemini 3.7 Flash<\/a> on August 13, describing it as its \u201cmost intelligent workhorse model yet for coding and agents.\u201d<\/p>\n Google has followed a consistent pattern since AI Mode launched, starting with Gemini 3 Pro\u2019s introduction to AI Mode<\/a> in November, making it an option alongside the existing default model. The next month, Gemini 3 Flash became the default in AI Mode<\/a>, and then at I\/O in May, Google replaced it with Gemini 3.5 Flash<\/a> as the global default.<\/p>\n In February, Jeff Dean, then Google\u2019s chief scientist, said Flash models stay Search\u2019s production tier due to their lower latency and cost. In my article<\/a> I noted this is the cycle Google plans to continue for future model generations.<\/p>\n Why This Matters<\/p>\n Currently, the value is somewhat limited. To select 3.7 Flash, you\u2019ll need an AI Pro or Ultra subscription, which means it doesn\u2019t reach AI Mode\u2019s free tier.<\/p>\n The rollout provides a helpful tool rather than setting a new default. If you\u2019re tracking inconsistent AI Mode citations, you can now see if the model itself is causing the differences by running the same query across 3.7 Flash, Auto, and Pro, then comparing the results.<\/p>\n Looking Ahead<\/p>\n