Google isn’t wasting any time. Just weeks after shipping Gemini 3.7 Flash, the company has already rolled out its successor, Gemini 3.8 Flash. The pitch is simple: the new model ‘works harder’ by running more reasoning steps on complex tasks and calling tools iteratively. That’s great for software engineering and autonomous agents, but it comes with a catch that developers will feel immediately in their cloud bills.
On paper, the per-token pricing hasn’t budged—$0.75 per million input tokens and $3.75 per million output tokens. But Google is warning that the model might burn through more tokens to hit peak performance, especially at higher effort levels. Artificial Analysis ran the numbers and found the real-world cost is up roughly 40% compared to 3.7 Flash, driven by a 30% jump in output tokens per task and more turns on agentic evaluations. If you want to keep costs down, Google says 3.7 Flash is still there for you.
Early reactions are enthusiastic, at least from people who care about code. Aigora.ai CEO John Ennis compared it to Anthropic’s top-tier models, claiming you get ‘Opus 5 coding quality but at a fraction of the cost and super fast.’ He thinks it’ll be particularly useful for things like generating Remotion videos. The benchmarks back up some of that swagger: Gemini 3.8 Flash beats its predecessor and other frontier models on DeepSWE v1.1, including Anthropic’s Fable 5, which also got a price cut on cached data earlier this week.
There’s a security angle here too. Google released Gemini 3.8 Flash Cyber alongside its new Fairwind Program, which is limited to governments and ‘trusted partners.’ The program’s 650 members include CrowdStrike and the Center for Internet Security. They get access to the Cyber model and CodeMender, an agent that autonomously finds and fixes vulnerabilities. The model also ships with safeguards against CBRN and cyber offense misuse. For regular users, Gemini 3.8 Flash is available now with Google AI Pro or Ultra subscriptions.
Google is clearly betting that developers will accept higher token consumption in exchange for better reasoning. But with Anthropic slashing cached-data pricing on Fable 5 just days ago, the cost-per-task battle is heating up. The question isn’t whether 3.8 Flash is good—it’s whether ‘good enough’ at a lower price still wins the market.
