Google's Gemini 3.8 Flash: One writer's favorite among four models

When Google offers several AI models to choose from, it can be hard to know which one to use. Writer Sanya Shah of How-To Geek tried four Gemini models and ended up with Gemini 3.8 Flash. It is her personal experience and not a systematic test.

The everyday favorite

Shah tested 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and 4 Argon. She picked 3.8 Flash and stuck with it after trying all four.

She describes the model as balanced and capable of handling multistep requests. She uses it for travel planning, research and comparisons, including an itinerary for Finland. According to Shah, this says a lot: "With Flash, I don't have to break my requests into smaller questions."

She considers Flash-Lite too simple for complex planning tasks and Pro unnecessary for everyday use. She sees Pro as a model for deeper analysis, such as comparing smartphones.

Gemini 4 Argon is the most capable of the models tested and is available only to AI Pro or Ultra subscribers. Shah finds it overkill for her needs and says: "I don't see enough reason to switch."

What Google says

Google released Gemini 3.8 Flash on September 2, 2026, along with 3.8 Flash Cyber, which is aimed at cybersecurity. According to Google, 3.8 Flash is significantly better than its predecessor, 3.7 Flash, and often approaches more expensive top-tier models. Google says it beats larger competitors on several benchmarks (standardized tests): DeepSWE v1.1, Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. These are the manufacturer's own claims.

The model is available in the Gemini app, AI Mode in Google Search and Google Sheets. Developers can use it in Google AI Studio, Android Studio and Antigravity, and businesses can get it in Gemini Enterprise.

Pricing and limitations

For developers using the model via API, there is an introductory price of $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. Tokens are the small chunks of text that AI models work with. Standard pricing applies from January 1, 2027.

According to Computing for Geeks, prices will double to $1.50 and $7.50 from that date. This has not been confirmed by Google, which only says that standard pricing begins then.

Computing for Geeks also reports that 3.8 Flash leads on 8 of 14 benchmarks against Claude Opus 5 and GPT-5.6, but scores clearly lower on open-ended agent tasks: 59 percent versus 75.4 percent on a computer use test. The figures come from a lesser-known third-party site and have not been further verified.

Developers should also know that the "minimal" setting for thinking level has been removed from version 3.7 onward. Requests that use it return an HTTP 400 error.