Google has announced the introduction of its latest model, "Gemini 3.8," which features improved reasoning and coding performance. This release includes two variants: "Gemini 3.8 Flash" and "Gemini 3.8 Flash Cyber."
According to Google, Gemini 3.8 Flash maintains the same speed and low cost as 3.7 Flash while providing improved reasoning and coding capabilities. For complex tasks, it is designed to process information more meticulously by executing additional reasoning steps or iteratively calling tools.
In benchmarks, Gemini 3.8 Flash has demonstrated high performance across multiple domains. It showed the ability to autonomously solve complex engineering problems in DeepSWE v1.1, and it outperformed 3.7 Flash and other frontier models in specialized domain benchmarks such as Vals Finance Agent V2 and Harvey's Legal Agent Benchmark.
Meanwhile, Gemini 3.8 Flash Cyber is a model specialized for the cybersecurity domain. It is currently being provided to select trusted defenders through the Fairwind Program. Google stated that it demonstrates frontier-level performance in autonomous vulnerability discovery in benchmarks such as CyberGym.
--- Sources:
Primary source: Gemini 3.8 Flash and 3.8 Flash Cyber (Hacker News Frontpage, 2026-09-03)