Google opens Gemini 3.8 Flash at half price and keeps the cyber version for defenders vetted into Fairwind
Google released Gemini 3.8 Flash with a cybersecurity twin, listing input at 0.75 dollars and output at 3.75 dollars per million tokens until December 31, 2026, after which both prices double. It is the third Flash model in six weeks, while Gemini 3.5 Pro and Gemini 4 have still not been released. The cyber version carries no price at all: defenders reach it only through Fairwind, an application-based programme Google says has more than 650 partners.
Artificial Intelligence··Night
The price holds at half until December 31 and doubles on January 1
Google released Gemini 3.8 Flash in two versions, a general reasoning and coding model and a cybersecurity variant called 3.8 Flash Cyber. Input lists at 0.75 dollars and output at 3.75 dollars per million tokens through December 31, 2026, and on January 1, 2027 the same list moves to 1.50 dollars for input and 7.50 dollars for output. Even after that rise the model stays far cheaper per token than the top models from OpenAI and Anthropic. It arrived three weeks after Gemini 3.7 Flash and is the third Flash release in six weeks, while the frontier models Gemini 3.5 Pro and Gemini 4 have still not been released. Developers reach it through Google AI Studio, Android Studio, the Gemini API, Google Antigravity and Stitch, enterprises through Gemini Enterprise, and subscribers through the Gemini app, AI Mode in Search and Google Sheets.[1], [3]
Google's own scores land just below Opus 5, and an independent index moves up three points
Google's own numbers put the model at 73.7 per cent on DeepSWE v1.1, a benchmark for long-horizon software engineering, just below Claude Opus 5 at 74.0 per cent and above GPT-5.6 Sol at 72.7 per cent and the previous 3.7 Flash at 65.3 per cent. On the company's other tests it reports 54.9 per cent on HLE-Verified and 47.2 per cent pass@1 on CWE-Bench patching, against 47.8 per cent for what it calls the leading frontier model at higher cost. No independent measurement of those results has been published. The benchmarking platform Artificial Analysis scores the model 59 on its Intelligence Index, three points above 3.7 Flash at 56. Google says part of the gain comes from the model working harder, running extra reasoning steps and calling tools iteratively, which raises token consumption and takes back part of the lower per-token price; where compute efficiency matters most the company points users to lower reasoning levels or the still-supported 3.7 Flash.[3], [1]
The cyber version is off the price list and goes through the Fairwind gate
The cyber variant does not appear on the price list at all. Defenders reach it only through Fairwind, an application-based programme that pairs 3.8 Flash Cyber with CodeMender, which Google says can find vulnerabilities and produce deployment-ready patches within minutes. Eligibility is limited to national cyber authorities, critical-infrastructure operators in health care, telecommunications, energy and finance, core platform companies and Google Cloud security partners; participants must confine access to their own incident-response teams and enforce multi-factor authentication. Google names Palo Alto Networks, Snowflake and Wiz among more than 650 partners, and puts its philanthropic side above 100 million dollars through Google.org, including 36 million dollars for 35 cyber clinics serving more than 1,250 hospitals, schools and utilities. The company reports a large gain in prompt-injection robustness for the 3.8 family as measured by Gray Swan, and says Flash Cyber passes 70 per cent success on a real-world evaluation spanning 20 programming languages. The partner count, the minutes-to-patch claim and those evaluation figures are the company's own, and no independent verification has been published.[2], [1]
