Quick answer
Google released Gemini 3.8 Flash on September 2, 2026 as a generally available model for coding, autonomous agents, and complex workflows. The API model ID is gemini-3.8-flash, with a one-million-token context window, up to 64,000 output tokens, and low, medium, or high thinking levels. Google also introduced Gemini 3.8 Flash Cyber, but access to that security-focused variant is restricted to vetted defenders through the Fairwind Program. Neither exact model is currently verified in Chat AI's model directory.
Gemini 3.8 Flash is generally available; the Cyber model is not
Google says Gemini 3.8 Flash is generally available and ready for production use through the Gemini API. Gemini 3.8 Flash Cyber uses the same foundational intelligence but is tuned for vulnerability discovery and automated patching. The Cyber variant is available only to approved government authorities, critical-infrastructure operators, software maintainers, and other trusted defenders through Google's Fairwind Program.
Sources: Google, Google AI for Developers, Google DeepMind
The standard model has a 1M-token context and three thinking levels
Developers can call the standard model with the ID gemini-3.8-flash. Google documents a one-million-token context window, a maximum output of 64,000 tokens, and low, medium, or high thinking levels, with medium as the default. Minimal thinking is not supported. Google cautions that the model may use more tokens on difficult tasks because it takes additional reasoning steps, calls tools iteratively, and checks its work.
Sources: Google AI for Developers
Introductory API pricing ends on December 31
Google lists introductory Gemini 3.8 Flash pricing of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. Starting January 1, 2027, the documented rates rise to $1.50 per million input tokens and $7.50 per million output tokens. These are API rates for the standard Flash model; they do not describe consumer subscription prices or Fairwind access terms.
Sources: Google, Google AI for Developers
Google reports stronger reasoning and defensive-security results
Google reports a 54.9% score for Gemini 3.8 Flash on HLE-Verified. For the restricted Cyber model, Google's Fairwind page reports 86.2% pass@1 on CyberGym, while the release post reports 47.2% pass@1 on the externally run CWE-Bench patching benchmark. These are provider-published results and do not guarantee performance on a particular repository, agent harness, or security review. Teams should reproduce relevant tests with their own code, tools, budgets, and human validation.
Sources: Google, Google DeepMind
Fairwind combines the Cyber model with access controls and CodeMender
Google says more than 650 partners participate in Fairwind. Approved teams agree to operational controls such as limiting use to internal security, incident-response, or penetration-testing staff and requiring protections including multi-factor authentication. Fairwind partners can also use Gemini 3.8 Flash Cyber with CodeMender, Google's agent harness for finding, validating, and patching software vulnerabilities. Enrollment is limited and requires an application; the announcement does not establish public API access.
Sources: Google DeepMind
Migration requires more than changing the model name
Google's migration checklist tells API users to replace the model ID, move from thinking_budget to thinking_level, remove deprecated sampling fields such as temperature, top_p, and top_k, and use server-side previous_interaction_id for multi-turn conversations. Function-calling applications should also review Google's updated payload and call-ID requirements. Test these changes in a staging environment before routing production traffic to the new model.
Sources: Google AI for Developers
Gemini 3.8 Flash availability in Chat AI has not been verified
Google says the standard 3.8 Flash model is available through Google AI Studio, the Gemini API, Gemini Enterprise, the Gemini app for Google AI Pro and Ultra subscribers, AI Mode in Search, and Gemini in Google Sheets. The exact Gemini 3.8 Flash and Gemini 3.8 Flash Cyber names do not appear in Chat AI's current verified model directory, so this article does not claim that either model is available in Chat AI.
Frequently asked questions
What readers usually ask
Is Gemini 3.8 Flash generally available?
Yes. Google describes gemini-3.8-flash as generally available and ready for production use through the Gemini API.
What context window does Gemini 3.8 Flash support?
Google documents a one-million-token context window and a maximum output of 64,000 tokens.
How much does the Gemini 3.8 Flash API cost?
Google lists introductory rates of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. The documented rates double on January 1, 2027.
Can anyone use Gemini 3.8 Flash Cyber?
No. Google says the Cyber variant is restricted to vetted defenders in its Fairwind Program, including selected government, critical-infrastructure, and software-security organizations.
Does Gemini 3.8 Flash support adjustable reasoning?
Yes. The API supports low, medium, and high thinking levels. Medium is the default, and minimal is not supported.
Is Gemini 3.8 Flash available in Chat AI?
The exact model is not listed in Chat AI's current verified model directory, so its availability in Chat AI is unverified.
Evidence