source: Google DeepMind: Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
level: technical
google deepmind introduced gemini 3.8 flash and gemini 3.8 flash cyber, two new models built on the same core intelligence. the flash variant is positioned as a workhorse for software engineering and agentic tasks, while the cyber variant focuses on vulnerability detection and automated patching. both are available at the same introductory price as the previous 3.7 flash: $0.75 per million input tokens and $3.75 per million output tokens.
gemini 3.8 flash shows strong results on benchmarks like deepswe v1.1, where it outperforms larger frontier models in long-horizon software engineering. it scores 54.9% on hle-verified, indicating multi-step reasoning ability. the cyber model achieves 47.2% pass@1 on cwe-bench, close to a leading frontier model at 47.8% but at lower cost. in internal tests, it found a critical vulnerability in under two hours, a task that usually takes months.
the cyber model is available only through the fairwind program for trusted defenders, reflecting safety concerns. google reports using it internally, with chrome security finding 2.6 times more correct patches than larger commercial models. the flash model is accessible via google ai studio, gemini api, and consumer apps. pricing will increase after december 31, 2026, to $1.50 per million input tokens and $7.50 per million output tokens.
why it matters: these models offer frontier-level coding and security capabilities at a fraction of typical costs, making advanced ai more accessible for enterprise and defensive use.
source: Google DeepMind: Introducing Gemini 3.8 Flash and 3.8 Flash Cyber