Google has officially launched Gemini 3.8 Flash on September 2, 2026, just hours after a report revealed that the new AI model could arrive this week. Internally referred to as “Skimaki,” the model is designed to improve coding, agentic workflows and complex multi-step reasoning while retaining the speed and lower cost associated with Google’s Flash series.
The launch is significant because Google is positioning Gemini 3.8 Flash as its most capable Flash model yet, while the company continues to compete aggressively with OpenAI and Anthropic in coding and AI-agent workloads. Google says the model is available today to developers and to Google AI Pro and Ultra subscribers in supported products.
What is Gemini 3.8 Flash?
Gemini 3.8 Flash is Google’s newest Flash-series AI model, built specifically around coding, agentic tasks and complex reasoning.
Google describes it as its “most intelligent workhorse model” and says it delivers significant improvements over Gemini 3.7 Flash without sacrificing the speed and cost advantages of the Flash family.
Unlike a major flagship model designed primarily to maximize raw intelligence, Flash models are intended to offer a balance between capability, speed and computing cost. That makes Gemini 3.8 Flash particularly relevant for developers building AI agents and software tools that need to make repeated model calls.
Why was Gemini 3.8 Flash called “Skimaki”?
“Skimaki” was reportedly the internal codename used for Gemini 3.8 Flash.
Before Google's official announcement, The Wall Street Journal reported that employees were testing a model called Gemini 3.8 Flash internally and referred to it as Skimaki. The report said the model could be released as early as September 2 and that some Google engineers preferred its performance to Anthropic's Opus model during internal testing in Google's Jetski coding platform.
That report has now effectively been overtaken by Google's announcement: Skimaki is Gemini 3.8 Flash, and the model has officially launched.
What can Gemini 3.8 Flash do?
Google is focusing Gemini 3.8 Flash heavily on long-horizon coding and autonomous agents.
The company says the model performs substantially better than Gemini 3.7 Flash on software engineering, agentic tasks and specialized multi-step reasoning. Google also says the model can repeatedly reason, call tools and refine its work on complex tasks instead of simply producing a response in one pass.
Some of Google's demonstrations include:
- Building an interactive 3D game using Google Antigravity and a simple prompt.
- Creating a playable DOS-style version of Google Maps.
- Generating scientific visualizations using real US Geological Survey datasets.
- Creating an interactive 3D hardware anatomy visualizer with Three.js.
- Handling long-horizon software engineering tasks autonomously.
These examples point toward Google's broader strategy of making Gemini useful not just as a chatbot, but as an agent capable of carrying out multi-step tasks.
How much does Gemini 3.8 Flash cost?
For developers using the Gemini API, Google has announced an introductory price of $0.75 per million input tokens and $3.75 per million output tokens. The introductory pricing is scheduled to expire on December 31, 2026. From January 1, 2027, Google says the prices will increase to $1.50 per million input tokens and $7.50 per million output tokens.
| Gemini 3.8 Flash API pricing | Price |
|---|---|
| Input tokens | $0.75 per 1M |
| Output tokens | $3.75 per 1M |
| Introductory pricing ends | December 31, 2026 |
| From January 1, 2027 | $1.50 input / $7.50 output per 1M |
This pricing is especially important for developers running AI agents, where a single task can involve many model calls.
How much better is Gemini 3.8 Flash at coding?
Coding is arguably the biggest story behind this release.
Google says Gemini 3.8 Flash outperforms most larger frontier models on DeepSWE v1.1, a benchmark focused on long-horizon software engineering. The company also says it performs better than Gemini 3.7 Flash and other frontier models on selected finance and legal agent benchmarks.
The model achieved 54.9% on HLE-Verified, according to Google, demonstrating its ability to work through multi-step questions spanning STEM, humanities and professional domains.
However, these are Google-reported benchmark results and should not automatically be interpreted as proof that Gemini 3.8 Flash is universally better than every competing model.
Is Gemini 3.8 Flash better than ChatGPT or Claude?
It is too early to declare an overall winner.
The WSJ report said Google employees preferred Gemini 3.8 Flash to Anthropic's Opus model in internal testing on Jetski. Google has separately published benchmark results showing strong performance from 3.8 Flash on coding and agentic tasks.
But Gemini 3.8 Flash is not necessarily designed to beat every OpenAI or Anthropic model at every task. Its biggest selling point is the combination of coding ability, reasoning, speed and cost.
| Area | Gemini 3.8 Flash |
|---|---|
| Coding | Major focus |
| AI agents | Major focus |
| Multi-step reasoning | Improved |
| Speed | Flash-series focus |
| API price | $0.75 input / $3.75 output per 1M tokens initially |
| Consumer availability | Google AI Pro and Ultra subscribers |
| Developer access | Gemini API, AI Studio and other Google tools |
The more meaningful comparison will come from independent testing after the model has been widely evaluated.
What is Gemini 3.8 Flash Cyber?
Google has also introduced Gemini 3.8 Flash Cyber, a specialized version focused on cybersecurity.
It is designed to detect vulnerabilities and help automate security patching. Unlike the standard 3.8 Flash model, the Cyber version is being provided through Google's Fairwind Program to trusted defenders such as government authorities, critical infrastructure operators and software maintainers.
Google says the model achieved a success rate above 70% on one of its internal vulnerability-discovery evaluations covering codebases across 20 programming languages.
The company also says its Chrome Security team found that 3.8 Flash Cyber produced 2.6 times more correct vulnerability patches than the best commercial models it compared against. These figures are Google's own testing results, rather than an independent industry-wide assessment.
Where can you use Gemini 3.8 Flash?
Google says Gemini 3.8 Flash is available from today across several of its AI products.
Developers can access it through the Gemini API, Google AI Studio, Android Studio and Google Antigravity. Enterprises can use it through Gemini Enterprise, while consumers can access it through Google AI Pro and Ultra subscriptions in the Gemini app, AI Mode in Google Search and Gemini in Google Sheets.
Google’s official Gemini 3.8 Flash announcement
Why is this launch important for Google?
The timing matters almost as much as the model itself.
Google has been releasing Flash models rapidly. Gemini 3.6 Flash arrived in July, followed by Gemini 3.7 Flash only a few weeks later. Google now says Gemini 3.8 is its third Flash release in six weeks.
The company is also trying to strengthen its position in coding and autonomous AI agents, two areas where OpenAI and Anthropic have attracted significant attention.
That makes Gemini 3.8 Flash less of a routine model refresh and more of a strategic push toward faster, cheaper and increasingly autonomous AI systems.
What about Gemini 4?
Gemini 3.8 Flash should not be confused with Google's next major flagship generation.
The pre-launch reporting around 3.8 Flash indicated that Google's next flagship model, Gemini 4, was still undergoing post-training. That means Gemini 3.8 Flash is better understood as a fast-moving Flash-series release rather than Google's next full flagship generation.
For users, this distinction matters: Google is effectively improving the “workhorse” side of Gemini while continuing development on its larger flagship models.
What does Gemini 3.8 Flash mean for AI users?
For everyday Gemini users, the most noticeable improvements may eventually come through better coding, stronger multi-step reasoning and more capable AI agents rather than a completely different chatbot experience.
For developers, the implications are potentially bigger. A faster and relatively inexpensive model that can reason through long tasks and repeatedly use tools could make autonomous coding agents and other AI workflows cheaper to operate at scale.
The launch also shows how quickly Google's Gemini lineup is evolving. Instead of waiting for one major annual model release, Google is now iterating through multiple Flash versions in a matter of weeks.
FAQ
What is the new Gemini model called?
The new model is Gemini 3.8 Flash. It was reportedly known internally by the codename “Skimaki” before Google's official announcement.
When did Gemini 3.8 Flash launch?
Google officially announced Gemini 3.8 Flash on September 2, 2026, after reports earlier in the day said the model could launch this week.
Is Gemini 3.8 Flash free?
Google's announcement lists consumer access through Google AI Pro and Ultra subscriptions. Developers can access the model through Google's API at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens.
Is Gemini 3.8 Flash better than ChatGPT?
There is no universal winner yet. Google reports strong coding and reasoning results, while the WSJ reported positive internal comparisons with Anthropic's Opus; independent testing will be needed for a broader comparison with OpenAI and Anthropic models.
Our Take
Gemini 3.8 Flash is interesting not simply because it is a new Gemini model, but because Google is pushing the Flash family toward serious agentic work. The combination of coding improvements, iterative reasoning and comparatively low API pricing could matter more to developers than another chatbot benchmark victory.
The bigger question now is whether Google's rapid Flash releases can translate into sustained advantages in real-world coding and AI-agent workflows. That will become clearer as developers and independent evaluators spend more time with the model.