On 2 September Google introduced Gemini 3.8 Flash and 3.8 Flash Cyber - the third Flash release in six weeks. The regular model costs the same as 3.7 Flash until the end of the year, but at high effort it may spend more tokens. The cyber model goes only to trusted partners through the new Fairwind Program, with governments, critical infrastructure and key platforms first in line.
- The introductory price stays at $0.75 per million input and $3.75 per million output tokens until the end of 2026; per Google, 54.9 per cent on HLE-Verified.
- Google says plainly that 3.8 Flash works harder: it takes extra steps and may use more tokens.
- Fairwind pairs 3.8 Flash Cyber with CodeMender to find and fix vulnerabilities; more than 650 partners, access only for internal security teams.
The same price per token. More tokens per task. The two things sit in the same Google announcement, and they have to be read together.
Gemini 3.8 Flash arrives three weeks after 3.7, at the same price. In the same text Google writes that the new model works harder: it takes extra reasoning steps, calls tools iteratively and sometimes spends more tokens, especially at high effort.
Price per token is no longer price per task
It is a small but important shift in the arithmetic. Until now you compared models from a table: this much per million in, this much out. When the new model thinks longer at the same price, the table stops telling the whole truth. Your bill depends on how many tokens it spends on your task, and you only find that out by running it.
The cyber half is more interesting politically. Google makes the move Anthropic and OpenAI made the day before: the strong cyber model goes only to vetted customers. Six hundred and fifty partners is a lot. It is still a list.
Using Flash through the API? Switch to 3.8 at low effort and compare a week's bill against 3.7. Not per million tokens. Per task.