Researchers from Nanjing University, SenseTime, and other institutions have released OpenMobile-2, a project showing that open-source smartphone agents can match the performance of proprietary models when trained on enough app variety.
App diversity outweighs data volume
Training on real applications is difficult because logins, purchases, and state management make the process expensive and hard to repeat. To solve this, the team rebuilt 35 popular apps, including Taobao, JD, DiDi, Meituan, Slack, and Google Drive. Combined with existing simulations and apps running in an Android emulator, the practice environment covers more than 110 applications.
Human reviewers compared the rebuilt apps to the originals and kept refining them until workflows and interfaces matched. Orders, budgets, and travel plans persist across sessions, and data can pass between apps via the clipboard. Every simulation resets to its starting state automatically, while the agent only sees screenshots and has no access to internal app data, mimicking how a person would actually use them.
From this setup, the team generated about 11,600 example workflows and more than 2,000 practice tasks with automated success checks. Google’s Gemini 3.1 Pro served as the teacher model that produced the examples, meaning OpenMobile-2 is effectively learning from a closed system.
The researchers held the amount of data constant and only increased the number of apps the agent practiced on. This pushed the success rate on unseen apps from 60.7 to 67.5 percent. With the same data budget, app variety beats extra examples from a handful of apps.
Performance on real devices leads the pack
Both models show clear gains on established benchmarks. On AndroidWorld, the larger model jumps from 67.1 to 79.9 points, while the smaller one goes from 54.3 to 74. The predecessor OpenMobile-1 scored 64.7 on the same test. On the harder MobileWorld benchmark, which covers longer tasks spanning multiple apps, the larger model improves from 35.9 to 50.4 points.
The biggest jump shows up on real Android devices, where the larger model climbs from 31.9 to 59.6 points on SPA-Bench. The researchers say that is the best score among open models.
Reinforcement learning on top brought only limited gains, mainly in environments that resembled the training setup. On AndroidWorld, the larger model actually scored slightly worse afterward, landing at 79.9 instead of 80.8 points.
Direct commands let agents skip the menu maze
An experimental part of the study goes beyond tapping and swiping. More than 50 of the simulated apps offer direct commands like “add to cart” so the agent can skip navigating through menus, and the agent freely switches between both approaches depending on what is available.
On its own, though, this does not help much because the models tend to ignore the shortcuts without targeted training. Once they have been trained on them, the agents use them well, solving up to 10.3 percentage points more tasks while needing about five fewer steps. The researchers say this ability also transfers to tools the models never encountered during training.
During the later reinforcement learning phase, tool use got a bonus factor whenever a run included at least one direct command, though the reward still depended on whether the task was actually completed. Penalties kicked in when the agent reported a success that was not real. These interfaces only exist in the simulated apps for now, and the researchers call the approach experimental themselves.
Some results come from a custom benchmark the team built with 215 tasks, nearly half of which span multiple apps. Task completion is not judged by a language model but by a program that compares the app’s stored data before and after each run. The authors also checked for overlap between training and test tasks and found only minor similarities.
Microsoft Research had already classified API agents as more reliable and GUI agents as more flexible in a 2025 study, pointing to mobile apps as a typical case where external interfaces do not exist. Oppo takes a different route with its open agent X-OmniClaw, which runs on real devices and replays previously shown paths in apps like Taobao or Meituan via deeplinks instead of clicking through menus again. Among closed models, Google has built smartphone control directly into Gemini 3.5 Flash.
What it means
For people making things, this means open-source tools can now handle complex, multi-step tasks on phones without needing a closed system to train them. The main requirement is access to a diverse set of simulated environments rather than massive datasets from a single provider. However, the direct command features described remain experimental and are currently limited to the simulation environment, so they are not yet available for general use on actual devices.
