Google has once again updated its “Android Bench” rankings for the best AI models for Android app development, with a bunch of new “open-weight” models as well as more details on the tokens used and cost of using these models.
One thing that large language models have gotten really good at is coding, with their ability to aid in the development of apps and other software projects also leading to the rise of “vibe coding.” Earlier this year, Google published a new benchmark ranking that showcased the “best” AI models for Android app development, taking into account common Android development tasks as well as how these models handle best practices.
When the “Android Bench” first debuted, Gemini 3.1 Pro led the pack, and OpenAI’s GPT 5.4 later tied for the top slot.
According to Google, GPT 5.5 is currently the best AI model for Android app development, beating out GPT 5.4 and Gemini 3.1 Pro by a little under 2%.
Are you using an AI models for Android app development?
Google has once again updated its “Android Bench” rankings for the best AI models for Android app development, with a bunch of new “open-weight” models as well as more details on the tokens used and cost of using these models.
One thing that large language models have gotten really good at is coding, with their ability to aid in the development of apps and other software projects also leading to the rise of “vibe coding.” Earlier this year, Google published a new benchmark ranking that showcased the “best” AI models for Android app development, taking into account common Android development tasks as well as how these models handle best practices.
When the “Android Bench” first debuted, Gemini 3.1 Pro led the pack, and OpenAI’s GPT 5.4 later tied for the top slot.
As of the May 18, 2026 update, there’s a new king in town. According to Google, GPT 5.5 is currently the best AI model for Android app development, beating out GPT 5.4 and Gemini 3.1 Pro by a little under 2%.
Advertisement - scroll for more content
But this latest update also puts things into perspective much better, as Google now shows the average latency, total tokens used, and the average cost of using each AI model. Google details how it arrived at each metric in documentation around the benchmark.
Average Latency: Time taken to solve 100 tasks across 10 runs
Time taken to solve 100 tasks across 10 runs Average Total Tokens: Token consumption for a full benchmark run across 10 runs
Token consumption for a full benchmark run across 10 runs Average Cost: Cost per benchmark run at the time of testing in US dollars
With that in mind, though, we can see that while GPT 5.5 is a bit stronger, it costs over twice as much to perform the same function as Gemini 3.1 Pro.
Here are the top ten models according to Google, including the new data (as of May 21, 2026):
Model Score Avg Latency Avg Total Tokens Avg Cost New: GPT 5.5 74 15.5 64.5 $133.9 GPT 5.4 72.4 21.2 64.2 $91.7 Gemini 3.1 Pro Preview 72.4 11.5 75.4 $49.0 New: Claude Opus 4.7 68.7 11.6 90.0 $124.3 GPT 5.3 Codex 67.7 11.2 71.4 $42.6 Claude Opus 4.6 66.6 9.9 69.5 $84.4 GPT 5.2 Codex 62.5 24.3 124.4 $121.9 Claude Opus 4.5 61.9 12.5 79.8 $102.5 Gemini 3 Pro Preview 60.4 9.8 117.0 $63.7 New: GLM 5.1 59.7 33.4 80.2 $46.7
As mentioned, there are more open-weight models now in the rankings, including Gemma, Qwen, DeepSeek, MiMo, and more. Of these, GLM 5.1 scored the highest, followed by Kimi K2.6.
You can see the full rankings on Google’s website.
Google continues to update the “Android Bench” on a roughly monthly basis. With Gemini 3.5 Pro coming soon and 3.5 Flash already live, it’ll be interesting to see if Google’s own models can tackle the lead OpenAI has now picked up.
Are you using an AI models for Android app development? If so, which one?
More on Android:
Follow Ben: Twitter/X, Threads, Bluesky, and Instagram
