AI Prices Just Moved Both Ways: What to Check Before December
By Adam, AI Engineer · 31 October 2026 · 3 min read
AI prices do not only go down. OpenAI has launched GPT-6.1 Terra at half the price of its predecessor, while Google's pricing page shows two popular Gemini models doubling in price within the next two months. If your product runs on one of these models, your bill could change without a single line of your code changing.
What changed
These are standard list prices per million tokens, checked on 31 October 2026 on the OpenAI and Google pricing pages.
| Model | Input | Output | Change |
|---|---|---|---|
| GPT-5.6 Terra | $2.00 | $12.00 | Older model |
| GPT-6.1 Terra | $1.00 | $5.00 | New, available now |
| Gemini 4 Argon | $2.00 | $10.00 | $4.00 / $20.00 from 1 Dec 2026 |
| Gemini 3.8 Flash | $0.75 | $3.75 | $1.50 / $7.50 from 1 Jan 2027 |
What it means in money
Take an example assistant handling 100,000 requests a month, with 3,000 input tokens and 400 output tokens per request. That is 300 million input tokens and 40 million output tokens, with no caching.
- GPT-5.6 Terra costs $1,080 a month. GPT-6.1 Terra costs $500.
- Gemini 4 Argon costs $1,000 a month now and $2,000 from 1 December.
- Gemini 3.8 Flash costs $375 a month now and $750 from 1 January.
A lower price does not prove a model suits your work, so these figures tell you what to test, not what to choose.
What to do this month
- Find out which models you use. Include features inside vendor products if your contract passes on usage costs.
- Recalculate your bill at the new prices using your real token numbers in our AI cost calculator.
- Test before switching. Run 50 to 100 real tasks through the cheaper option and compare quality, not only the price.
- Make switching easy. Keep model names in one setting rather than across your code, so the next price change is a small job.
The bigger lesson
Treat AI model prices like any other supplier price. They change, sometimes on short notice, so a product's margins should not rely on one model staying cheap. Before you sign off on an AI feature, check whether it still makes money if the model price doubles. If it does not, fix that before launch.