Google ships Gemini 3.8 Flash, third release in six weeks

Google released Gemini 3.8 Flash on September 2, 2026, the third Flash-tier model shipped in six weeks and the fastest cadence yet for the line. Google positions it as its most capable Flash model for agentic coding, built to work through longer chains of tool calls and reasoning steps on complex tasks instead of returning a fast single-pass answer.

On Google's own benchmark results, the model outperforms larger frontier models on DeepSWE v1.1, a software engineering benchmark, and beats its predecessor, Gemini 3.7 Flash, on Vals AI's Finance Agent V2 and Harvey's Legal Agent Benchmark. It scores 54.9% on HLE-Verified, a multi-step reasoning test spanning STEM and professional domains.

Pricing starts at $0.75 per million input tokens and $3.75 per million output tokens, an introductory rate that holds through December 31, 2026. Google hasn't said what the model will cost after that window closes. The model is generally available now through the Gemini API, Google AI Studio, Antigravity, Gemini Enterprise, and the Gemini app.

Google also introduced Gemini 3.8 Flash Cyber, a separate variant for trusted testers in a new Fairwind Program aimed at autonomous vulnerability discovery. For developers already routing agent workloads through Gemini's Flash tier, the pricing window is the practical deadline: budget for a token-cost increase once the introductory rate expires, or lock in current usage patterns before it does.