Google has introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, two new models in its Flash family. Gemini 3.8 Flash targets coding, software engineering, agentic workflows and multi-step reasoning, while the Cyber variant is designed for vulnerability discovery and automated patching for trusted defenders. The release gives developers, paid Gemini users and Gemini Enterprise customers different routes to use the general model, with a separate access program for the cybersecurity model.
The announcement, published September 2, 2026, is Google’s third Flash release in six weeks and follows Gemini 3.7 Flash. In Google’s official Gemini 3.8 Flash and Flash Cyber announcement, the company calls 3.8 Flash its most intelligent workhorse model and says it delivers substantial improvements over 3.7 Flash on long-horizon reasoning and multi-step tasks.
For businesses, the most immediate significance is practical rather than abstract. A faster or more capable model is useful when it can reliably support real work that spans several steps, such as interpreting a request, working with code or documents, checking intermediate results and producing an output that a person can review. Google positions Gemini 3.8 Flash around precisely those coding and agent-oriented workflows.
What Gemini 3.8 Flash changes
Google says Gemini 3.8 Flash shares a foundational intelligence with the Cyber variant and is accelerated by agentic loops. In this context, agentic workflows are tasks in which a model can work through a sequence of actions and decisions instead of providing only a single response. That makes model quality on longer reasoning chains particularly relevant for development and operational automation.
The stated improvements are centered on:
- Coding and software engineering, including work handled through developer tools.
- Agentic workflows that require a model to complete connected steps.
- Long-horizon reasoning, where a task needs sustained progress across multiple stages.
- Multi-step tasks, where an answer depends on processing and checking more than one instruction or input.
Google has made 3.8 Flash available through the Gemini API in Google AI Studio and through Android Studio. It also highlights Google Antigravity workflows for agent-first development and Stitch for UI generation. Consumers with Google AI Pro or Ultra subscriptions can access the model in the Gemini app, AI Mode in Google Search and Gemini in Google Sheets. Gemini Enterprise is another access route for enterprise customers.
That distribution matters because it separates experimentation from implementation. Teams can test a use case through Google’s consumer and developer surfaces, but production use will depend on whether the API, development environment or enterprise route fits their existing processes. The announcement confirms the access points, but it does not provide implementation guidance for every type of workflow or business system.
| Area | Gemini 3.8 Flash | Gemini 3.8 Flash Cyber |
|---|---|---|
| Primary focus | Coding, software engineering, agentic workflows and multi-step reasoning | Vulnerability discovery and automated patching |
| Intended users | Developers, consumers and Gemini Enterprise customers through specified Google products | Trusted defenders |
| Access route | Gemini API, Google AI Studio, Android Studio, Gemini apps and Gemini Enterprise | Fairwind Program |
| Pricing disclosed in the announcement | $0.75 per million input tokens and $3.75 per million output tokens initially | Not specified in the supplied announcement summary |
Pricing and timing
Gemini 3.8 Flash launches at the same introductory API price as Gemini 3.7 Flash: $0.75 per million input tokens and $3.75 per million output tokens. Google says this introductory pricing runs through December 31, 2026. From January 1, 2027, the listed rates rise to $1.50 per million input tokens and $7.50 per million output tokens, respectively.
The scheduled increase is important for teams assessing recurring AI costs. A prototype can be inexpensive while still masking the cost of frequent production use, especially where a workflow sends large documents, codebases or repeated task context to a model. Testing should therefore measure actual input and output volume, not just the quality of a few individual prompts.
Flash Cyber is for a restricted defender audience
Gemini 3.8 Flash Cyber is not presented as a broadly available security assistant. Google describes it as its most capable cybersecurity model, with frontier-level performance for vulnerability discovery and automated patching. Access is available through a new Fairwind Program for trusted defenders, with priority given to government authorities, critical infrastructure operators and software maintainers.
This distinction is central to the rollout. Companies should not assume that access to Gemini 3.8 Flash includes Flash Cyber. The Cyber model has a dedicated defender-access path, reflecting the dual-use nature of advanced cybersecurity capabilities.
Google also says the models include safeguards against misuse in CBRN and cyber-offense domains, operate within its Frontier Safety Framework and improve prompt-injection robustness. Those measures are relevant for organizations considering agent-style systems, where untrusted content can be introduced through documents, websites, tickets or other external inputs. The announcement does not eliminate the need to review how a workflow handles permissions, sensitive information and human approval of consequential outputs.
For companies building software or automating internal work, Gemini 3.8 Flash may be most useful when applied to a bounded, measurable process. Examples could include drafting and reviewing code changes, classifying incoming requests, preparing a first response from approved information or helping staff move through a defined multi-step task. The model’s claimed reasoning gains should be tested against the specific task, data and quality threshold that matter to the business.
A capable model alone does not create an efficient process. Scalevise's AI workflow automation service can help turn model access into reliable workflows that reduce manual handling, connect the right business tools and keep people involved where review matters. The practical next step is to identify one repetitive, multi-step process, measure its current effort and assess whether Gemini 3.8 Flash can improve it without creating avoidable operational risk. Request an AI automation consultation with Scalevise.
Frequently Asked Questions
What is Gemini 3.8 Flash?
Gemini 3.8 Flash is a Google Flash-family model designed for coding, software engineering, agentic workflows, long-horizon reasoning and multi-step tasks. Google describes it as the most intelligent workhorse model in its suite.
Where can developers access Gemini 3.8 Flash?
Google says developers can access Gemini 3.8 Flash through the Gemini API in Google AI Studio and through Android Studio. Google also highlights Antigravity workflows and Stitch in connection with development workflows.
How much does the Gemini 3.8 Flash API cost?
The introductory price is $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. Google lists prices of $1.50 per million input tokens and $7.50 per million output tokens from January 1, 2027.
Who can use Gemini 3.8 Flash Cyber?
Gemini 3.8 Flash Cyber is available to trusted defenders through Google’s Fairwind Program. Google prioritizes government authorities, critical infrastructure operators and software maintainers.
Conclusion
Google’s Gemini 3.8 Flash rollout expands access to a model positioned for more demanding coding and multi-step work, while Flash Cyber creates a specialized channel for cybersecurity defenders. The announced API pricing and broad product availability make 3.8 Flash a concrete option for testing AI-supported workflows. Its business value will depend on disciplined evaluation of task quality, token use, workflow design and appropriate human review.
Top comments (0)