AI News·13 Nov 2026·via Dev Community·Covered by 7 sources

    Google switches gemini-2.5-pro and gemini-2.5-flash to a newer model on 1 December 2026: what to change

    Google will transition its `gemini-2.5-pro` and `gemini-2.5-flash` AI models to `gemini-3.8-flash` and `gemini-3.5-flash-lite`, respectively, starting December 1, 2026. While API requests to the older models will continue to be answered, the underlying AI models will change, requiring developers to prepare for potential shifts in output.

    The upcoming model transition for Google's Gemini APIs on December 1, 2026, involves more than a simple name update. Developers using `gemini-2.5-pro` and `gemini-2.5-flash` will find their requests answered by newer `gemini-3.8-flash` and `gemini-3.5-flash-lite` models. This shift means that while API calls will still function, the new models may produce different outputs in terms of format, length, tone, and JSON parsing. Google's documentation indicates specific parameter changes are required for Gemini 3 models, such as replacing `thinking_budget` with `thinking_level` and adjusting for function calling.

    To manage this transition effectively, developers are advised to proactively identify all instances where the older model IDs are used within their codebases. The article suggests command-line tools like `npx @aidriftline/driftline` or `grep` to locate these references, though it notes that names set in environment variables, databases, or SDK defaults may not be found by such searches. This proactive identification is crucial to avoid unexpected behavior in applications once the switch occurs.

    Crucially, testing applications with the replacement models before the December 2026 deadline is recommended. Developers should run 10 to 20 real prompts from their logs on both the current and new models to compare responses. This allows for adjustments while the older models are still accessible for comparison. The article mentions Driftline Pro, a service that runs application prompts on both models in GitHub Actions to highlight changes, and a free Driftline GitHub Action, which flags model retirements in pull requests, as tools to aid this process.

    Share this article

    Go deeper
    Original reporting by Dev CommunityWe don't republish, read the full story →

    Related reading

    6 stories