AI API prices are not fixed. As of 19 October 2026, Google's Gemini pricing page lists scheduled increases that double some rates from December 2026 and January 2027, while Anthropic's page says a previously scheduled increase for one model will not occur. Treat every price as a dated input, and build review dates into your budget.
Examples from the public pages
From Google's Gemini API pricing page, read on 19 October 2026:
- Gemini 4 Argon: input $2.00 and output $10.00 per million tokens, rising to $4.00 and $20.00 from 1 December 2026.
- Gemini 3.8 Flash and Gemini 3.6 Flash: input $0.75 and output $3.75, rising to $1.50 and $7.50 from 1 January 2027.
From Anthropic's pricing documentation on the same date, a footnote says the $2 and $10 per million input and output pricing for Claude Sonnet 5, which ran through 31 August 2026, is now the standard price, and that the increase scheduled for 1 September 2026, to $3 and $15, will not occur.
The same Anthropic page also says Claude 4.7 and later models use a tokenizer that produces approximately 30 percent more tokens for the same text. That is a price change in effect even when the list price stays flat.
Older models also leave the menu. The page marks several earlier models as retired, except on some cloud platforms.
Why this matters for a budget
A fixed budget built on today's list price can be wrong by a factor of two next quarter. The change can also arrive through another door: a model retirement that forces a move, a tokenizer change, a context threshold, or a new default model.
A planning checklist
- Record the date and source URL next to every price in your budget sheet.
- Calendar a price review each month, and the dates of any announced changes.
- Model a high case in which your main model's price doubles. Know what you would do.
- Keep a second model tested and ready, so a price move does not become an emergency migration. See choose a model tier by cost and task.
- Abstract the model name in code and config, so switching is a setting, not a rewrite.
- Re-count tokens for any new model before you estimate its cost. See what is an AI token.
- Keep spend caps in place so a surprise shows up as a limit, not an invoice.
What not to do
Do not sign long commitments on the strength of one price page, and do not tell customers a fixed AI feature price unless you control the margin.
Quick check
Will prices go down?
Sometimes, and sometimes up. The pages read for this post show both a cancelled increase and scheduled increases. Plan for movement in either direction.
Where should I read the current price?
On the provider's own pricing page, on the day you make the decision.
Sources
Last reviewed: 19 October 2026.
