{"id":297,"date":"2026-11-04T10:14:35","date_gmt":"2026-11-04T10:14:35","guid":{"rendered":"https:\/\/scoy.ai\/guides\/google-reroutes-old-gemini-models\/"},"modified":"2026-11-04T10:14:35","modified_gmt":"2026-11-04T10:14:35","slug":"google-reroutes-old-gemini-models","status":"publish","type":"post","link":"https:\/\/scoy.ai\/guides\/google-reroutes-old-gemini-models\/","title":{"rendered":"On December 1, Google Starts Answering Gemini 2.5 Pro Calls With 3.8 Flash"},"content":{"rendered":"\n

If the chatbot that answers your customers’ emails runs on Google’s Gemini 2.5 Pro or 2.5 Flash, don’t wait for December 1 to find out which model it calls. That’s the day Google starts answering those calls with different models, billed at the new models’ prices. Do you know which one yours uses?<\/p>\n\n\n\n

Google’s release note gives the date and the new model’s name. It doesn’t price the swap, and it doesn’t say whether your old settings carry over. It also leaves out that Google’s own deprecations page said the 2.5 models “will continue to be served until further notice” as late as October 24.<\/p>\n\n\n\n

Google keeps two schedules for these models. Google Cloud’s page lists October 20. The Gemini API is how software asks Google’s model a question. Its page listed no date at all until this:<\/p>\n\n\n\n

Effective December 1, 2026, all API requests to gemini-2.5-pro will automatically route to gemini-3.8-flash and be billed at Gemini 3.8 Flash rates.<\/p>

Google, Gemini API release notes<\/a>, November 2, 2026<\/p><\/blockquote>\n\n\n\n

The same row of Google’s deprecations page, in two archived copies and live this morning:<\/p>\n\n\n\n

Date<\/th>Shutdown date listed for gemini-2.5-pro<\/th><\/tr><\/thead>
October 2, Internet Archive copy<\/a><\/td>No shutdown date announced<\/td><\/tr>
October 24, Internet Archive copy<\/a><\/td>No shutdown date announced<\/td><\/tr>
November 4, live page<\/a><\/td>December 1, 2026<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Flash got the same turn. Its calls go to Gemini 3.5 Flash-Lite, which Google’s pricing page calls a “cost-efficient model.” Google also says access to the 2.5 models is “limited to users who have actively used them in the past,” so this lands on tools that have been running a while. All of it comes from Google’s own pages. That makes it one source.<\/p>\n\n\n\n

Does the Swap Make Your Bill Smaller?<\/h2>\n\n\n\n

In December, yes. From January, it depends on what you send in versus what comes back.<\/p>\n\n\n\n

Rerouted calls get billed at the new model’s rates. For the Pro model’s replacement, Google’s pricing page<\/a> shows introductory rates through December 31 that double on January 1. I worked out three kinds of workload from that page on November 4. The prices are the standard paid tier, for prompts under 200,000 tokens. A token is a small chunk of text, about a word.<\/p>\n\n\n\n

What goes in and what comes back<\/th>Old Pro model<\/th>New model, December<\/th>New model, from January 1<\/th><\/tr><\/thead>
Long documents in, short answers out (1,000,000 in, 20,000 out)<\/td>$1.450<\/td>$0.825<\/td>$1.650<\/td><\/tr>
Ten in for every one out (1,000,000 in, 100,000 out)<\/td>$2.250<\/td>$1.125<\/td>$2.250<\/td><\/tr>
Chat and drafting (100,000 in, 100,000 out)<\/td>$1.125<\/td>$0.450<\/td>$0.900<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

Ten to one is the line. Send more than ten times as much text in as comes back, and January costs more than the old model did. Send less, and it costs less. Is your work mostly long documents and short answers? Then you’re above the line.<\/p>\n\n\n\n

I can’t tell you where your tool sits. The model’s thinking is billed as output, and I haven’t measured how much the new one uses.<\/p>\n\n\n\n

The Flash move leaves the price alone. Text in and out costs the same on both models, $0.30 and $2.50 per million tokens, on that same page. What changes is the model. I haven’t compared the two on real work, so I can’t say whether your answers change.<\/p>\n\n\n\n

Will Your Old Settings Survive the Swap?<\/h2>\n\n\n\n

Google doesn’t say. Its notes warn of “breaking API changes” on both new models, and Google’s list of changes for developers<\/a> names them:<\/p>\n\n\n\n