Brief IA
›

Google Launches Gemini 3.8 Flash and a Limited Cyber Edition

AI Tools·Tom Levy·

Google Launches Gemini 3.8 Flash and a Limited Cyber Edition

Google Launches Gemini 3.8 Flash and a Limited Cyber Edition
Key Takeaways
1Google releases Gemini 3.8 Flash and a Cyber edition, with prices maintained until December 31, 2026
2Prices will double on January 1, 2027, for token entry and exit
3The publisher claims better performance in reasoning and coding, but increased token consumption at high effort
4Access to the Cyber variant is through the Fairwind program, reserved for certain players
Why it matters : Google continues to develop its Gemini range with an extended pricing offer and differentiated safeguards for cybersecurity.

Google adds a dual variant to its Gemini 3.8 Flash lineup, with prices frozen until December 31, 2026, and a Cyber version available by application. The company describes 3.8 as its best iteration for reasoning and coding, while warning of higher token consumption at significant effort levels.

Extended Pricing and Controlled Access for the Cyber Edition

Google is extending the introductory pricing conditions for Gemini 3.8 Flash until December 31, 2026, at $0.75 per million input tokens and $3.75 per million output tokens. On January 1, 2027, these amounts are set to double to $1.50 and $7.50. The general model includes protections against CBRN and cyber-offensive uses as part of the Frontier Safety Framework, while the Cyber variant applies more permissive cybersecurity mitigations, which justifies restricted access. Access to Gemini 3.8 Flash Cyber is limited to selected participants through the Fairwind program, targeting public authorities, critical infrastructure operators, and software maintenance managers. This access restriction extends the approach already adopted for Gemini 3.5 Flash Cyber, which was reserved for governments.

Claimed Performance and Contribution from Cybersecurity Work

Google presents Gemini 3.8 Flash as its best model for reasoning and coding. The demonstration highlights the DeepSWE v1.1 benchmark, focused on long software engineering tasks, where the model reportedly outperforms most larger frontier systems, according to the company. Google attributes part of the progress in coding and reasoning of the common foundation to work conducted in cybersecurity, described as particularly demanding. The Gemini 3.8 Flash Cyber variant aims at vulnerability detection and automated remediation, with a priority, as asserted by Google, given to remediation rather than offensive capabilities.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

More Iterative Operation and Token Cost Management

Gemini 3.8 Flash multiplies reasoning steps and tool calls iteratively, which can increase token consumption when the effort level is high. For workloads where computational cost is the main constraint, Google recommends lowering this effort level or sticking with 3.7 Flash, which remains supported. The company also indicates that speed and cost are expected to be equivalent to 3.7 Flash for this new version.

Availability for Developers, Businesses, and Subscribers

Gemini 3.8 Flash is accessible to developers via Google Antigravity, the Gemini API in Google AI Studio, Android Studio, and Stitch. Businesses can access it through the Gemini Enterprise offering. Google AI Pro and Ultra subscribers can use it in the Gemini app, the AI Mode of Search, and Google Sheets. This version, the third Flash iteration in six weeks, comes with a Cyber variant, both relying on the same technical foundation; the general version aims at a broad audience.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.