The Cheap AI Model You Just Wired In Doubles in Price on January 1

What shipped. September 1: Anthropic released Claude Fable 5.1 and cut cache-read pricing by 75%. September 2: Google released Gemini 3.8 Flash at $0.75 per million input tokens and $3.75 per million output tokens, labeled introductory pricing through December 31, 2026. The listed price from January 1, 2027 is $1.50 and $7.50. Google also shipped a security variant, Gemini 3.8 Flash Cyber, held back from general release behind a vetted "Fairwind" access program that prioritizes government, critical infrastructure, and software maintainers.

The so-what for operators. Introductory pricing is now a standard launch move, and it carries an expiration date almost nobody writes down. If you or your developer picked a model this month because it was cheap, your run rate can double on a Wednesday in January with no warning email. The second signal is worth noting too: the strongest defensive models are gated, while general-purpose models keep getting cheaper and more capable for everyone, including whoever is writing the phishing email aimed at your controller.

The 20-minute expiry check.

1. List every place AI costs you money. Seats (ChatGPT, Claude, Gemini), API keys buried inside automations (Zapier, Make, n8n, anything a developer built you), and AI features bundled into software you already pay for.
2. For each one, write down the model name and today's price. If a developer built it, ask them one line: "Which model does this call, at what rate, and is that rate introductory?"
3. Flag anything labeled introductory, promotional, preview, or beta pricing. Those are the ones that reset.
4. Put each reset date on your calendar 30 days early, with current monthly spend in the invite body so future-you has something to compare against.
5. Cap the damage now. Set a hard monthly spend limit in every API console. A capped key that stops is a bad afternoon. An uncapped key that runs is a bad quarter.

Paste this to build the ledger:

You are a fractional CTO. I run a [business type] with [X] people. Below is every AI tool and automation we pay for, with what I know about each. Build a table with these columns: Tool, What it does for us, Model or plan, Current monthly cost, Pricing status (standard / introductory / unknown), Reset or renewal date, Internal owner, What breaks if it stops. Mark every row where pricing status is introductory or unknown as ACTION, and for each of those give me one specific question to ask the vendor or the developer. Then tell me which three rows carry the most dollar risk if the price doubles. Here is my list: [paste]

Keep invoices and internal cost data in a private or business AI workspace rather than a public chat.

Sources: Google: Gemini 3.8 Flash and 3.8 Flash Cyber · Google: the Fairwind Program