Google launched Gemini 3.8 Flash on Wednesday in an effort to reclaim its reputation as a top tier model maker. Doubts about Google and the frontier model race grew in early August when Google DeepMind CEO Demis Hassabis stepped down to become chairman, and Koray Kavukcuoglu took over as senior vice president.
Google CEO Sundar Pichai failed to convince skeptics after the company missed its promised June release for Gemini 3.5 Pro. Modest benchmark metrics for Gemini 3.5 Flash also hurt confidence, as Chinese open-weight models overshadowed Google's speed and cost efficiency with higher intelligence scores.
Google has maintained a rapid iteration cycle by releasing Gemini 3.8 Flash as its fourth Flash model in as many months. The new model approaches the intelligence of higher cost frontier models while retaining performance and affordability.
That pricing will change in the new year when the introductory rate of $0.75 per million input tokens and $3.75 per million output tokens doubles. Tulsee Doshi and Raluca Ada Popa wrote in a blog post that Gemini 3.8 Flash delivers substantial gains over version 3.7.
The authors stated that the model often approaches the performance of expensive frontier models. They added that on DeepSWE v1.1 for long-horizon software engineering, the model outperforms larger frontier models in solving complex engineering problems at a fraction of the cost.
Gemini 3.8 Flash configured to high reasoning scores 59 on the Artificial Analysis Intelligence Index, which is a three point gain over its predecessor. That score matches GPT-5.6 Sol on extra high reasoning and Grok 4.6 on medium reasoning.
Other models rank higher on current intelligence tables based on nine benchmarks. GLM-5.3 and Kimi K3 score 60, Grok 4.6 on high reasoning scores 61, GPT-5.6 Sol on max reasoning scores 61, Claude Opus 5 scores 63, and Claude Fable 5.1 scores 66.
Artificial Analysis calls Gemini 3.8 Flash the cheapest model at its intelligence level at $0.58 per Intelligence Index task. Anthropic's flagship Fable 5.1 costs about six times more at $3.76 per task.
The new model costs roughly 40 percent more per task than its predecessor because it generates more tokens and executes more turns as an agent. Doshi and Popa explained that the model works harder on complex tasks by executing extra reasoning steps and calling tools iteratively.
The model also shows improvements on benchmarks for enterprise knowledge work including Vals Finance Agent V2, Harvey's Legal Agent Benchmark, and HLE-Verified. Developers can access Gemini 3.8 Flash through Google Antigravity, the Gemini API in Google AI Studio and Android Studio, and Stitch.
Enterprises can use Gemini Enterprise, while consumers with Google AI Pro and Ultra subscriptions can access the model via the Gemini app, AI Search, and Google Sheets. Google also launched the Fairwind Program to give governments, critical infrastructure organizations, and software maintainers access to Gemini 3.8 Flash Cyber for vulnerability hunting and remediation.
