Google updates Android Bench rankings, Gemini 4 Argon lags behind Claude and GPT models

·by Henderson·Gemini
Google updates Android Bench rankings, Gemini 4 Argon lags behind Claude and GPT models

Google has updated its Android Bench list to show the performance of AI models in assisting Android coding. According to the latest update, Gemini 4 Argon's ranking is significantly lower than that of the latest Claude and GPT models. Android Bench was launched earlier this year to help Android developers quickly understand which AI models are best suited for Android coding assistance, considering AI usage related to Android coding and development best practices.

As Google continues to test and rank new models, average cost has also been added as a benchmark to help developers make the best decisions for their needs.

In the update on October 26, Google added the inference levels used by each model and included results for OpenAI's GPT-6 Sol, GPT-6.1 Sol, Claude Sonnet 5, and Claude Sonnet 5.5, along with Google's own Gemini 4 Argon. Currently, the top rank is held by Claude Sonnet 5.5, with a pass rate of 40% and a completion rate of 80.9%.

Claude Opus 5.5, added in early October, appears in second place.

Claude Sonnet 5.5 excels in AI model rankings

OpenAI's GPT-6.1 Sol may be the most notable addition here, with a pass rate of 29.3% and a completion rate of 82.8%, but at a cost of US$99.50 (approximately HK$777), compared to Claude's US$1,574.90 (approximately HK$12,277) for the top rank. In contrast, Google's Gemini 4 Argon debuted with relatively disappointing results, ranking fifth and trailing behind other leading models.

Its pass rate is only 21.3%, with a completion rate of 81.8% and an average cost of US$267.70 (approximately HK$2,086).

Shortly after Google announced Gemini 4 Argon, some internal voices indicated that the model performed poorly on certain coding tasks, although Google expressed support for its capabilities.

H
About the author
Henderson