Google Releases Gemini 3.8 Flash and 3.8 Flash Cyber · Chronopin

Google Releases Gemini 3.8 Flash and 3.8 Flash Cyber · Chronopin

Google introduced Gemini 3.8 Flash on 2 September 2026, three weeks after 3.7 Flash and its third Flash release in six weeks, calling it its best reasoning and coding model yet at the same speed and low cost as 3.7^[1]
It launched alongside Gemini 3.8 Flash Cyber, Google's most capable cybersecurity model for vulnerability detection and automated patching, available only to trusted defenders through the new Fairwind Program because it ships with more permissive cyber mitigations^[1]
The Fairwind Program pairs Gemini 3.8 Flash Cyber with Google's CodeMender harness for governments and national cyber authorities, critical infrastructure operators and core technology platforms, and Google says it has more than 650 participating partners^[4]
Gemini 3.8 Flash reached Google AI Pro and Ultra subscribers in the Gemini app, AI Mode in Search and Gemini in Google Sheets, developers in Google Antigravity, AI Studio and the Gemini API, and enterprises in Gemini Enterprise^[1][5]
The launch came as Google shares ended a four-month losing streak, their longest since 2015; the stock rose 0.6% on the day, and D.A. Davidson's Gil Luria said the model keeps Google in the race but probably leaves it a distant third in the enterprise market^[6]
Google followed on 15 September with Gemini 3.8 Live and 3.8 Live Extended Thinking, its live voice models^[3]

Notable features

Introductory price of $0.75 per million input tokens and $3.75 per million output tokens, the same as 3.7 Flash, until 31 December 2026; from 1 January 2027 it rises to $1.50 and $7.50^[1]
A 1,048,576-token context window and 65,536 maximum output tokens, text, image, audio and video input with text output, and LOW, MEDIUM (default) and HIGH thinking levels, with no MINIMAL level^[2]
Benchmarks in Google's table: 73.7% on DeepSWE v1.1 against 65.3% for 3.7 Flash, 54.9% on HLE-Verified, 61.4% on Vals Finance Agent v2 and 89.4% on Terminal-bench 2.1^[1]
Designed to work harder, taking extra reasoning steps and calling tools iteratively, so it may use more tokens at higher effort levels^[1]
Knowledge cutoff of March 2026 for some domains and January 2025 for others^[5]
Flash Cyber exceeds a 70% success rate on Google's internal 20-language vulnerability benchmark, scores 47.2% pass@1 on CWE-Bench, and gave the Chrome Security team 2.6 times more correct patches than much larger commercial models^[1]