Google Gemini 3.8 Flash is designed for “multitasking”.

Gemini 3.6 Flash, Google’s previous smartest AI model, was released at the end of July. The tech giant wasted no time in repeating it, launching Gemini 3.7 Flash a few weeks later. Now, just three weeks later, the company is introducing the successor to Flash 3.7.

Google just posted Gemini 3.8 Flash and this time focus on delivering next-generation intelligence with two different versions for agency workflows and cybersecurity.

The new models mark Google’s third Flash release in six weeks, with the new Frontier models nowhere to be seen. The new models are Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. The former is the company’s “smartest operating model” and the latter is “the most capable cybersecurity model.”

Gemini 3.8 Flash is primarily aimed at developers building agency applications and workflows, but it is also available to regular Google AI Pro and Ultra subscribers through the Gemini app.

According to Google, Gemini 3.8 Flash outperforms the largest frontier models when it comes to the DeepSWE v1.1 benchmark. The benchmark measures the ability to autonomously solve complex engineering problems, and Gemini 3.8 Flash achieved 73.7% of it. This is slightly below the 74% score of the Claude Opus 5 and well above the Gemini 3.7 Flash, Claude Sonnet 5, GPT-5.6 Sol and GPT-5.6 Terra.

Gemini 3.8 Flash also includes financial analysis in Vals Finance Agent v2, complex legal workflows in Harvey’s Legal Agent Benchmark, information synthesis in CharXiv Reasoning, agent terminal coding in Terminal-bench 2.1, long video understanding, and more. including outperforming all of the aforementioned models in several other benchmarks.

Google explained that “3.8 Flash works harder,” meaning performance increases. In complex tasks, it shows great care, if necessary, the model performs additional proof steps. This means that the model may sometimes use more symbols to improve performance.

Gemini 3.8 Flash starts at $0.75 per million input tokens and $3.75 per million output tokens. These are introductory pricing that will be revised down to $1.50 and $7.50, respectively. Regardless, the model should still be cheaper than counterparts like the Claude Opus 5 and GPT-5.6 Sol.

Gemini 3.8 Flash Cyber, on the other hand, is not available to everyone. Access to it is limited through the new Fairwind program, which allows trusted government agencies, as well as critical infrastructure operators and software providers, to use the new model.

On the industry standard benchmark for detecting vulnerabilities in C/C++, the model scored 86.2%, beating its predecessor Gemini 3.5 Flash Cyber ​​and larger models such as Mythos 5 and GPT-5.5-Cyber. Similarly, in real-world vulnerability detection involving more than 20 programming languages, the model achieved a success rate of 71%, significantly higher than Gemini 3.7 Flash’s 58.9% and Gemini 3.5 Flash Cyber’s 46.6%.

Elsewhere, what makes the model stand out is its automated patching capability. Gemini 3.8 Flash equips cyber defenders with the ability to patch vulnerabilities. According to the company, the model produced 2.6 times the correct patches for vulnerabilities in Chrome, “much larger than the best commercial models.”

Most regular users will never actually interact with the new model, but the improvements it offers to defenders and developers can have the indirect effect of making your everyday tech safer and better.

Gemini 3.8 Flash is available now for developers, Enterprise customers, and Google AI Pro and Ultra subscribers. Availability of Gemini 3.8 Flash Cyber ​​is, as previously mentioned, limited to members of the Fairwind program.