Google releases Gemini 3.8 Flash, its third new AI model in six weeks
A standard version for coding and everyday tasks, plus a security-focused variant, both priced at half the regular rate until the end of 2025.
Key points
- Google launched Gemini 3.8 Flash, its third Flash model release in six weeks.
- The model ships in two versions: a general-purpose workhorse and a cybersecurity-tuned edition called Gemini 3.8 Flash Cyber.
- Introductory pricing is $0.75 per million input tokens (small chunks of text fed to the model) and $3.75 per million output tokens, roughly half the standard rate.
- Google has not released a frontier-level Gemini Pro model since early 2026, and the promised Gemini 3.5 Pro looks increasingly unlikely to appear.
Google has launched Gemini 3.8 Flash, its best reasoning and coding model yet, by its own account. Three Flash releases in six weeks is a pace worth pausing on.
What does Gemini 3.8 Flash actually do?
It handles two jobs. The standard Gemini 3.8 Flash is a general-purpose model built for everything from writing code to agentic tasks, software that carries out multi-step jobs on its own without a human clicking through each step. Google calls it a workhorse for developer tools and AI assistants.
The second version, Gemini 3.8 Flash Cyber, runs on the same underlying technology but has been fine-tuned for cybersecurity work: spotting vulnerabilities in software and suggesting fixes. It's aimed at security teams and developers who need an AI that understands attack patterns. We covered the Cyber model's Fairwind Program on 2 September, when Google announced it would write and test security patches autonomously.
What does this cost developers?
For now, cheap. Google is offering API access, the programming interface that lets developers connect their apps to the model, at an introductory rate through the end of 2025: $0.75 per million input tokens and $3.75 per million output tokens. A token is roughly three-quarters of a word, so a million covers a very large body of text.
The standard rate after the promotional period rises to $1.50 and $7.50 per million tokens respectively. Google pitched the same introductory structure for the 3.7 Flash release a couple of weeks ago. Those prices probably reflect competitive pressure: rival AI labs have recently cut token costs to keep increasingly wary businesses engaged, and Google's matching that move. It's also worth noting that new models will likely arrive before the promotional window closes, as Anthropic's cheaper Claude pricing showed on 1 September.
Where is Gemini Pro?
Nobody at Google is answering that directly. The company hasn't shipped a frontier-level Gemini Pro model since early 2026. Frontier models are the biggest, most capable systems a lab produces; Pro sits above Flash in raw power. A Gemini 3.5 Pro was promised, but three Flash launches in six weeks suggests Google may have quietly shelved it in favour of iterating on the faster, cheaper Flash line instead.
For businesses waiting on a more powerful Pro upgrade before committing to Google's AI stack, it's time to reassess.
Common questions
Will this affect Gemini inside Google's apps?
Not immediately in a visible way. Flash models power many everyday Gemini features already, and 3.8 Flash is faster and more capable than its predecessors, so responses may feel snappier.
Is the half-price deal worth locking in now?
The introductory rate lasts through the end of 2025. Developers building tools that call the API heavily could save meaningfully, though another model will probably arrive before the pricing changes anyway.
