Reports published on 2 September 2026 say Google is preparing a new Google Gemini AI model with a stronger emphasis on coding. Those accounts describe an upcoming release. They are not a public product announcement from Google.

The company has not confirmed the full set of details circulating about the model or when it would become available.

What reports say about the new Gemini model

Coverage says the system is known inside Google as Skimaki and is expected to be called Gemini 3.8 Flash. Some reports add that a release could come as early as this week.

Those points should be read as reporting, not as a confirmed schedule. Names used internally, a public product name, and a launch window can all change before an official announcement.

The Wall Street Journal has reported that internal testing of the upcoming model showed progress on coding work. Google has not publicly confirmed that testing, the results, or a release date.

Why a coding-focused Flash model would matter

AI-assisted programming has become one of the most competitive areas among large AI companies. OpenAI, Anthropic, and Google are all developing models meant to help developers write, debug, understand, and improve software.

A new model in Google’s faster Flash series would continue a strategy the company already uses: Gemini models that aim for quick responses while handling a high volume of tasks. That describes the Flash line in general. It does not confirm how any unreleased model would perform.

If Google does ship Gemini 3.8 Flash with the reported coding gains, developers would see another competitor enter a market already contested by Gemini, OpenAI, and Anthropic systems.

What remains unconfirmed

Google has not published specifications, pricing, or a launch date for this reported Google AI model. Until it does, the timing and the claimed improvements should be treated as unconfirmed.

For developers, the useful question will be how the model handles real programming work, not only benchmark charts. Generating reliable code, reading an existing codebase, finding bugs, and staying useful across longer tasks would decide whether a new Gemini model is actually helpful.

Google has been updating its AI models at a rapid pace. Another Gemini release, if it arrives, would add to that competition. For now, the reported launch remains a report.

Related policy coverage: the United States is separately pressing G20 governments toward a lighter approach to AI regulation.