Gemini 4 Argon spotted in A/B testing on Antigravity

The quiet appearance of Gemini 4 Argon in A/B tests suggests Google is gathering real-world feedback before a broader release. Users on X, particularly those with Jio-bundled Gemini Pro plans or sub-accounts in family groups, have reported receiving outputs they believe are from Argon, sometimes routed behind other model names. This method allows Google to test the model's performance and user experience in diverse conditions.

Initial user observations, such as those from @leo114119 and @auricxofficial, indicate a noticeable step up in output quality. The example of Argon generating more polished UI screens for a weather app in HTML, compared to Gemini 3.8 Flash, highlights an improvement in design aesthetics. This addresses earlier criticisms of Gemini models producing generic or less detailed outputs, potentially enhancing Argon's appeal for applications requiring visual sophistication.

While Google has published benchmark claims, including a 77.9% score on DeepSWE for software engineering tasks, these early user experiences offer a different perspective. The article notes that Google is testing various internal checkpoints, such as Barium-B for the public release and Carbon for internal employee use. This internal variation means that the specific version of Argon encountered by early testers might differ from the final public offering, making a comprehensive assessment challenging until wider access is granted.