Gemini 3.8 Flash, Explained Simply
Google shipped Gemini 3.8 Flash on September 2, 2026, its newest fast, lower-cost AI model, along with a specialised Cyber variant built for security work. We explain what actually changed and why it matters, without the benchmark jargon.
⚡ Quick facts
- Released: September 2, 2026
- Price: $0.75 per million input tokens and $3.75 per million output, rising to $1.50 and $7.50 on January 1, 2027
- Google's claim: better than 3.7 Flash at coding, agentic tasks and multi-step reasoning, with stronger prompt-injection defences
- Cyber variant: limited access through Google's Fairwind program; Google claims 2.6x more correct Chrome patches than larger models
What Gemini 3.8 Flash actually does
It's a multimodal model, meaning it can take in text, images, audio and video, with a 1 million token context window (roughly enough to process a long book in one go). Google calls it its "most intelligent workhorse model" and says it improves on 3.7 Flash in software engineering, agentic tasks and multi-step reasoning, Thurrott reported. It arrived about three weeks after 3.7 Flash, and Deccan Chronicle notes it "works harder" by using extra reasoning steps and tool calls, which can mean more tokens per answer at higher effort levels.
The video trick: watching on demand, not frame-by-frame
Most AI models process video by breaking every single frame into data upfront, which gets expensive fast for long videos. On September 1, 2026, Google added agentic video understanding to its Flash models: the model navigates a video's timeline and pulls just the frames, transcript snippets or audio it needs, similar to how a person scrubs through a video. OfficeChai reported the launch for 3.7 Flash, 3.6 Flash and 3.5 Flash-Lite with about 88% fewer tokens; Google has not said clearly whether 3.8 Flash gets the same mode, so check the API docs before relying on it. The same on-demand approach powers Gemini features that arrived in Android's September update.
How it compares
Google reports 54.9% on HLE-Verified and says 3.8 Flash outperforms 3.7 Flash and other frontier models on Vals Finance Agent V2 and Harvey's Legal Agent Benchmark, and most larger models on DeepSWE v1.1, though it did not publish every score. These are Google's own numbers; independent, side-by-side testing will tell the fuller story, but on paper this is a meaningful step up for Google's "fast and cheap" model tier.
Gemini 3.8 Flash Cyber: built for security work
Alongside the main model, Google released Gemini 3.8 Flash Cyber, tuned specifically for cybersecurity tasks like finding and patching vulnerabilities. Google says its Chrome Security team found it produced 2.6 times more correct patches for Chrome vulnerabilities than the best commercial models, even much larger ones; that is Google's own test, not an independent one. It is not on the public API: access goes through Google's Fairwind program for governments, critical infrastructure operators and core technology platforms, and Google says it vets every applicant (DataCamp has a summary).
Our take: what a real workload costs before and after January
The headline price matters less than the January change, so we worked through one example. Say a support team runs 10,000 conversations a day, each using about 3,000 input tokens and 500 output tokens:
- Daily tokens: 30 million in and 5 million out.
- Until December 31, 2026: 30 × $0.75 + 5 × $3.75 = $22.50 + $18.75 = $41.25 a day, about $1,240 a month.
- From January 1, 2027: 30 × $1.50 + 5 × $7.50 = $45 + $37.50 = $82.50 a day, about $2,480 a month.
Three things to check before committing:
- Budget on the 2027 price, not the introductory one.
- Measure tokens per answer. Google says 3.8 Flash can use more reasoning steps; if each answer grows by a third, the saving over a bigger model shrinks.
- Cap the effort level for simple tasks and keep high effort for the steps that need it.
Pricing and where to use it
Through the API, Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, identical to 3.7 Flash's pricing. Google has said both prices double to $1.50 and $7.50 on January 1, 2027, so this is an introductory rate. Developers reach it through the Gemini API in Google AI Studio and Vertex AI. Consumers on Google AI Pro or Ultra get it in the Gemini app, AI Mode in Search and Gemini in Google Sheets.
Sources
- Thurrott: Google releases Gemini 3.8 Flash and Gemini 3.8 Flash Cyber
- Deccan Chronicle: Google launches Gemini 3.8 Flash with enhanced reasoning
- DataCamp: Gemini 3.8 Flash Cyber explained
- OfficeChai: agentic video understanding in Gemini Flash models
Frequently asked questions
Can I get access to Gemini 3.8 Flash Cyber?
Only through Google's Fairwind program, which is limited to governments and national cyber authorities, critical infrastructure operators and core technology platforms. Google says it vets every applicant; the Cyber model is not on the public Gemini API.
What is Gemini 3.8 Flash?
Google's newest fast, lower-cost multimodal AI model, released September 2, 2026.
What's new?
Google says it is better than 3.7 Flash at coding, agentic tasks and multi-step reasoning, and resists prompt injection better.
What is the Cyber variant?
A version tuned for cybersecurity tasks, available only through Google's vetted Fairwind program. Google claims 2.6x more correct Chrome patches than larger rivals.
How much does it cost?
$0.75 per million input tokens and $3.75 per million output tokens via the API until December 31, 2026, then $1.50 and $7.50.