Google releases Gemini 3.8 Flash and a cybersecurity variant

single source · 1 articles · release · confidence: high · first seen 2026-09-02 00:00 UTC

Google has released Gemini 3.8 Flash three weeks after 3.7 Flash, along with Gemini 3.8 Flash Cyber, a cybersecurity model available through the Fairwind Program for trusted defenders. The standard model costs $0.75 per million input tokens and $3.75 per million output tokens, the same as 3.7 Flash. Google reports 54.9% on HLE-Verified and says the model outperforms most larger frontier models on DeepSWE v1.1, with gains on Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. The improvements come from the model working harder: it uses extra reasoning steps and iterative tool calls, sometimes spending more tokens.

What this means for you

If you run Gemini 3.7 Flash, you can evaluate 3.8 Flash at the same per-token price, but expect higher token usage on complex tasks, so cost per request may rise. For automated security work, 3.8 Flash Cyber is gated by Google's Fairwind Program.

Key facts

  • ·Google DeepMind released Gemini 3.8 Flash on 2 September 2026, three weeks after Gemini 3.7 Flash. source
  • ·Gemini 3.8 Flash is priced at $0.75 per million input tokens and $3.75 per million output tokens, the same as 3.7 Flash. source
  • ·Gemini 3.8 Flash Cyber is available through the Fairwind Program for trusted defenders. source
  • ·Gemini 3.8 Flash scores 54.9% on HLE-Verified. source
  • ·Google says Gemini 3.8 Flash outperforms most larger frontier models on DeepSWE v1.1. source

What the sources say

  • Google DeepMind BlogAnnounces the benchmark scores, pricing, and gated cybersecurity variant for Gemini 3.8 Flash.

Sources

The original reporting. Follow these — they did the work.

← the wire