// AI NEWS · MODEL RELEASE

Google Launches Gemini 3.8 Flash and a Cybersecurity Variant — Its Third Flash Model in Six Weeks

Announced September 2, 2026, Gemini 3.8 Flash is a harder-working coding and agents model Google says tops larger rivals on some tasks, shipping alongside a defense-focused 3.8 Flash Cyber — at the same fast, cheap price as 3.7 Flash.

2026-09-02 · by Moe Ameen

What happened

On September 2, 2026, Google announced Gemini 3.8 Flash, describing it as a workhorse model built to "work harder" — putting in extra reasoning steps and more iterative tool use to push through complex, multi-step engineering and analysis tasks. It is the third release on Google's cost-efficient Flash tier in roughly six weeks, following Gemini 3.6 Flash in late July and Gemini 3.7 Flash on August 13. Like those, it is a text-and-reasoning model for coding, agents, and knowledge work — not an image or video generator, despite the "Flash" name it shares with Google's separate media models.

Google reports gains over 3.7 Flash and says the model outperforms several larger frontier models on some benchmarks while staying cheap. The company highlights autonomous engineering (a test it labels DeepSWE v1.1) plus stronger results in quantitative and professional domains, citing the Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. Google also shipped a specialized sibling, Gemini 3.8 Flash Cyber, tuned for defensive security — vulnerability discovery and automated patch generation — which Google reports reaches frontier-level results (for example around 47.2% pass@1 on a patching benchmark it calls CWE-Bench and a real-world vulnerability-discovery success rate above 70% across 20 programming languages). Access to the Cyber variant is limited to trusted defenders through Google's Fairwind Program. As always with launch-day figures, treat vendor benchmarks as Google's own until third parties test them.

Pricing matches 3.7 Flash's introductory offer: $0.75 per million input tokens and $3.75 per million output through December 31, 2026, then $1.50 per million input and $7.50 per million output starting January 1, 2027. Gemini 3.8 Flash is available in the Gemini app for Google AI Pro and Ultra subscribers, in AI Mode, in Gemini for Google Sheets, and to developers through the Gemini API, Google AI Studio, Android Studio, and Google Antigravity, plus Gemini Enterprise.

Why it matters for creators

  • The upgrade is diligence, not distribution. A model that "works harder" reasons through a research pile or a multi-step plan more thoroughly — but it still generates no video, no carousel, and no scheduled post.
  • Cheap drafting stays cheap. The introductory $0.75 / $3.75 per-million rate carries over from 3.7 Flash, so batch text and planning remain nearly free through year-end.
  • The wins are in coding, finance, and legal reasoning. That helps the upstream ideation-and-analysis stage of content work, not the production-and-publishing stage.
  • The Cyber variant is a defensive-security tool, gated to trusted defenders — not something in a creator workflow. Do not confuse the two releases.
  • The introductory price expires December 31, 2026 and roughly doubles. If you wire a pipeline to it, budget for the January 1, 2027 step-up.

How to act on this with Kompozy

The pitch on 3.8 Flash is that it is more diligent — it takes extra reasoning steps and chains more tool calls to finish a hard task. That is genuinely useful for the front of a content workflow: hand it a stack of sources and it will summarize, extract, and plan a week of angles more carefully than a lighter model. But diligence is not distribution. A more thorough plan is still just a plan — the model records no video, designs no carousel, writes no caption in your voice, and posts to nothing. That gap is structural to a language model, and it did not narrow with this release.

[Kompozy](/) is the layer that closes it. Take the researched angles or rough scripts you generate with 3.8 Flash and drop them in: Kompozy rewrites them in your voice through a [Persona Brief](/glossary/persona-brief), then generates the finished formats a model can't — captioned [Persona Shorts](/glossary/persona-shorts) and HeyGen avatar video, brand-exact carousels and quote cards via [HyperFrames](/glossary/hyperframes), photo posts, blog articles, and email newsletters — and schedules and publishes the whole set across the eight social platforms plus blog and email on [Autopilot](/glossary/autopilot). Pair the cheap, careful brain with the engine that turns its plan into a published content week.

Quick takeaways

  • Google announced Gemini 3.8 Flash on September 2, 2026 — its third Flash release in about six weeks — calling it a workhorse model built to "work harder" on complex coding and analysis.
  • Google reports gains over 3.7 Flash and says it beats some larger frontier models on tests like DeepSWE v1.1, the Vals Finance Agent V2, and Harvey's Legal Agent Benchmark (vendor benchmarks).
  • A sibling model, Gemini 3.8 Flash Cyber, targets defensive security — vulnerability discovery and automated patching — and is limited to trusted defenders via the Fairwind Program.
  • Pricing matches 3.7 Flash: introductory $0.75 / $3.75 per million input/output tokens through December 31, 2026, then $1.50 / $7.50 from January 1, 2027.
  • It is a text-and-reasoning model, not a media generator. Kompozy is what turns its drafts and plans into finished, published content across platforms.

Frequently asked questions

What is Gemini 3.8 Flash, and when did it launch?

Google announced it on September 2, 2026 — its third Flash release in roughly six weeks. It is a fast, cost-efficient "workhorse" model for coding, agents, and knowledge work that Google says "works harder" by taking extra reasoning steps and using tools iteratively. It is a text-and-reasoning model, not an image or video generator.

What is Gemini 3.8 Flash Cyber?

It is a specialized sibling of 3.8 Flash tuned for defensive cybersecurity — vulnerability discovery and automated patch generation, with capabilities prioritized toward defense over offense. Access is limited to trusted defenders through Google's Fairwind Program, so it is a security tool rather than a general or content model.

How much does Gemini 3.8 Flash cost, and can it make social posts?

It launched at an introductory $0.75 per million input tokens and $3.75 per million output through December 31, 2026, then $1.50 / $7.50 from January 1, 2027. It generates and reasons over text but produces no images, video, or scheduled posts — turning its drafts into finished, on-brand content across platforms is a separate job handled by a content engine like Kompozy.

Related news

← All AI news · Get started →