Google releases Gemini 3.8 Flash with iterative tool use and higher output token usage

Google released Gemini 3.8 Flash, describing it as a model that "works harder" than its predecessor by performing more reasoning steps and calling tools iteratively on complex tasks.

The company maintained the same introductory pricing as Gemini 3.7 Flash—$0.75 per million input tokens and $3.75 per million output tokens—but warned that the model might use more tokens to maximize performance, especially at higher effort levels.

Developers can continue using Gemini 3.7 Flash if they aim to minimize token usage.

Early third‑party analysis noted that while per‑token pricing is unchanged, output token usage per task increased by about 30% and more turns were observed in agentic evaluations, potentially raising effective costs.

Google claims the new model delivers "significant improvements" for software engineering and autonomous AI agents, citing better performance than its predecessor and other frontier models on the DeepSWE v1.1 software engineering benchmark, including Anthropic’s Fable 5.

Google also reports that the model outperformed competitors on the Vals Finance Agent V2 benchmark and Harvey’s Legal Agent benchmark.

Gemini 3.8 Flash includes safeguards against misuse in Chemical, Biological, Radiological, and Nuclear (CBRN) and cyber offense domains.

A variant, Gemini 3.8 Flash Cyber, launched alongside Google’s Fairwind Program, which is limited to governments and trusted partners and includes access to the CodeMender agent for autonomous vulnerability remediation.