Google has unleashed Gemini 3.8 Flash upon the world, a mere few weeks after its predecessor, because why let a perfectly good naming scheme gather dust? The company claims this new model "works harder" than Gemini 3.7 Flash - presumably by pulling all-nighters and doing extra reasoning steps on complex tasks while "calling tools iteratively," which sounds exhausting.

But here's the twist: despite keeping the same introductory pricing as 3.7 Flash - $0.75 per million input tokens and $3.75 per million output tokens - this industrious little model could still end up costing you more. Google warns, with the subtlety of a forehead slap, that "the model might use more tokens to maximize performance, especially at higher effort levels." So it's like a contractor who charges the same hourly rate but suddenly takes twice as long because he's really committed to doing a thorough job. Developers who prefer their AI to be more of a lazy genius can stick with Gemini 3.7 Flash to keep token usage down.

Early impressions are rolling in, and they're a mixed bag of praise and math. Artificial Analysis crunched the numbers and declared Gemini 3.8 Flash "the cheapest we've measured at this level of intelligence," but then dropped the bombshell: "This is up ~40% from Gemini 3.7 Flash despite unchanged per-token pricing, driven by a 30% increase in output tokens per task and more turns on agentic evaluations." So it's cheaper per IQ point, but you're paying more per task because it just can't stop chattering.

Aigora.ai CEO John Ennis is positively giddy, comparing Gemini 3.8 Flash to Anthropic's models: "Opus 5 coding quality but at a fraction of the cost and super fast." He adds, "This is going to be so awesome for things like making remotion videos." We'll take his word for it.

Google touts "significant improvements" for software engineering and autonomous AI agents. On the DeepSWE v1.1 software engineering benchmark, it outperformed its predecessor and other frontier models, including Anthropic's Fable 5 - which got an upgrade earlier this week that also promises stronger performance at a lower price, by cutting the price to use cached data. It's a regular arms race of AI one-upmanship.

The new model also claimed victory on the Vals Finance Agent V2 benchmark and Harvey's Legal Agent benchmark. But lest you think it's all fun and games, Google notes that Gemini 3.8 Flash "ships with safeguards against misuse in the domains of Chemical, Biological, Radiological, and Nuclear (CBRN) and cyber offense." Because even hardworking AI needs to be told not to build a bioweapon.

In a related move, Google released Gemini 3.8 Flash Cyber alongside its new Fairwind Program, which is strictly for governments and "trusted partners." The program, boasting 650 members including CrowdStrike and the Center for Internet Security, offers access to 3.8 Flash Cyber and Google's CodeMender agent, which can "autonomously find and fix vulnerabilities, protecting critical infrastructure, public services, and national security." Because nothing says "trusted partner" like a government-approved cyber AI.

Gemini 3.8 Flash is available now for consumers with a Google AI Pro or Ultra subscription, as well as for developers and enterprise users. So if you're ready for your AI to work harder (and possibly bill you more), it's time to upgrade.