OpenMobile-2: Building Versatile Mobile Agents with Scalable Environments and App-Native Tools
Authors: Chenyang Yan, Yingying Zhang, Kanzhi Cheng, Qiushi Sun, Zhengyuan Pan, Zheng Ma, Hang Yan, Yian Wang, Nuo Chen, Jialin Cao, Xingdong Gong, Wenpo Song, Han Chen, Zichen Ding, Fangzhi Xu, Shujian Huang, Xinyu Dai, Yichen Liu, Tiankuo Yao, Bo Wang, Ben Kao, Jianbing Zhang, Lewei Lu, Dahua Lin
Organizations: Nanjing University · SenseTime · The University of Hong Kong · Xi’an Jiaotong University · National University of Singapore · East China Normal University
We introduce OpenMobile-2, a near-frontier mobile agent with fully open training environments and recipes. We make three key advances: (1) Diverse environments with simulated commercial apps: We build MobileGym++, featuring 35 realistic, functionally rich commercial-style apps with cross-app workflows, while preserving full controllability for reset and verification. Together with newly configured apps in Android emulators, this yields a diverse playground spanning over 110 apps. (2) Open training data at scale: Building on this foundation, we curate nearly 12K mobile interaction trajectories for supervised fine-tuning and the largest open collection of verifiable RL training data for mobile agents, comprising over 2K executable tasks with automatic rewards. (3) Hybrid GUI and app-native tool use: We explore an experimental mobile-use setting where apps expose selected functionalities as app-native tools alongside their GUIs, allowing agents to interleave GUI actions and tool calls within a task. We implement this setting in MobileGym++ with over 300 carefully scoped tools across 50+ apps. Additionally, we introduce MobileGym++ Bench for complex, long-horizon commercial mobile scenarios, supporting both GUI-only and hybrid GUI-tool evaluation on a shared task suite. OpenMobile-2 performs competitively across established benchmarks, including AndroidWorld (79.9) and MobileWorld (50.4), while showing promising transfer to real-device mobile use, nearly doubling SPA-Bench performance from 31.9 to 59.6. All resources are available at https://os-copilot.github.io/OpenMobile2-Home/.
Figures & tables
Figure 1: OpenMobile-2 performs strongly across established mobile agent benchmarks and achieves leading results among open-source models on real-device evaluation.
Figure 2: (a) Application coverage of OpenMobile-2. Each square represents two applications; dark and light tiles denote simulator and emulator apps, and dots mark apps with native tools. (b) Hybrid GUI and app-native tool use. The same task can be completed through GUI-only interaction (top) or by interleaving GUI actions with app-native tool calls (bottom) over a shared app state.
Environment
Platform
Scale
Interaction
Full Controllability
# Comm. Apps
# Tasks
Cross-app
GUI
App-native Tools
AppWorld ( Trivedi et al., 2024 )
API
9
750
✓
✗
✗
✓
AndroidLab ( Xu et al., 2025 )
Emulator
5
138
✗
✓
✗
✗
AndroidWorld ( Rawles et al., 2025 )
Emulator
0
116
✓
✓
✗
✗
SPA-Bench ( Chen et al., 2025 )
Real-device
58
340
✓
✓
✗
✗
MobileWorld ( Kong et al., 2026 )
Emulator
4
201
✓
✓
✗
✗
Table 1: Comparison of mobile agent environments across scale, interaction, and controllability. ✓ denotes supported, ✗ unsupported, and ∼ indicates limited system level cross-app interaction support in MobileGym. # Comm. Apps and # Tasks denote the number of covered commercial apps and benchmark tasks, respectively; Detailed application statistics are provided in Appendix C . MobileGym++ provides diverse, high-fidelity commercial-style applications with full controllability, while introducing an experimental app-native tool interface.
Figure 3: App-level interaction richness. Each marker denotes one app, positioned by its interactive actions and page transitions; marker size reflects the number of Bench tasks using that app. Our new apps reach 1.48 × more actions and 1.91 × more transitions.
Figure 4: Diversity and coverage of OpenMobile-Data-v2. (a) Domain distribution across seven categories for emulator and simulator data. Gini coefficients quantify domain balance. (b) Pairwise cosine similarity among task goals. (c) Single-app and cross-app workflow composition.
Method
AndroidWorld
MobileWorld
MobileGym
MobileGym++
SPA-Bench
Pass@1 ↑
Pass@3 ↑
GUI
MCP
SR
PR
GUI
Hybrid
L3
Commercial & Closed-Weight Models
GPT-5.6-Sol
–
–
70.1
–
–
–
–
–
68.1*
Gemini-3.1-pro-preview
70.7
–
58.1
–
–
–
–
–
70.2*
MAI-UI-32B
73.3
–
36.2
30.0
–
–
–
–
–
MAI-UI-235B-A22B
76.7
–
39.7
37.5
–
–
–
–
–
Table 2: Main results on established mobile agent benchmarks and real-device evaluation. ∗ denotes results reproduced by us, and ± denotes the standard deviation across evaluation seeds.
Figure 5: Environment coverage and app-native tool use. (a) Broader coverage improves held-out AndroidWorld performance at a fixed 2K-trajectory budget. (b) App-native tools improve success and efficiency, with larger gains after Hybrid SFT.
Table 8
Appendix figures & tables9 assets
Supplementary material from the paper’s appendix.
Appendix
Hyperparameter
SFT
RL
Framework
Megatron-SWIFT
verl
Objective
Next-token prediction
DAPO (group-relative)
Trainable parameters
LLM (vision frozen)
LoRA
Optimizer
Adam ( β1=0.9 , β2=0.95 )
AdamW
Weight decay / gradient clip
0.1 / 1.0
0.1 / 1.0
Learning rate
5×10−6
1×10−6
Appendix
Table 5: Training hyperparameters. SFT values are shared by both runs; – indicates that a setting does not apply.
GUI-only run
Hybrid run
Source
Trajectories
Steps
Trajectories
Steps
Emulator
7,725
83,631
7,725
83,631
Simulator
2,144
22,857
3,918
34,951
Total
9,869
106,488
11,643
118,582
Appendix
Table 6: Trajectories in each SFT run. The hybrid run adds 1,774 simulator trajectories that interleave app-native tool calls, so the simulator count rises from 2,144 to 3,918.
Figure 6: Per-goal maximum cosine similarity. (a) SFT goals against the matched benchmark. (b) RL task goals against MobileGym++ Bench. Vertical lines mark each mean and the 0.7 threshold.
Benchmark
Train goals
Mean
Max
Above 0.7
AndroidWorld
7,718
0.434
0.892
1.11% (86)
MobileWorld
7,718
0.391
0.851
0.39% (30)
MobileGym
3,916
0.458
0.826
0.46% (18)
MobileGym++ Bench
3,916
0.504
0.861
2.25% (88)
Appendix
Table 7: Maximum cosine similarity of each SFT training goal to the matched benchmark. Above 0.7 counts goals whose maximum exceeds 0.7.
Domain
MobileGym++
Emulator
Lifestyle
18
19
Productivity
15
23
Entertainment
9
15
Shopping
8
1
Travel
5
3
Social
6
0
Appendix
Table 8: Application counts by domain in the two runtimes.
Table 10: The 62 Android emulator applications, grouped by domain. AndroidWorld and MiniWoB denote the task-support apps installed with the AndroidWorld suite.
Figure 8: MobileGym++ Bench composition. Left: tasks by the number of applications they touch. Right: task incidences by domain. Domain counts are not mutually exclusive.
Domain
Task incidences
Productivity
97
Finance
34
Shopping
53
Social
35
Entertainment
26
Lifestyle
46
Appendix
Table 11: MobileGym++ Bench task incidences by domain. A task that touches more than one domain is counted in each, so the column does not sum to 215.
Mobile agents powered by vision-language models have demonstrated impressive capabilities in automating mobile tasks, with recent leading models achieving a marked performance leap, e.g., nearly 70% success on AndroidWorld. However, these systems keep their training data closed and remain opaque about their task and trajectory synthesis recipes. We present OpenMobile, an open-source framework that synthesizes high-quality task instructions and agent trajectories, with two key components: (1) The first is a scalable task synthesis pipeline that constructs a global environment memory from exploration, then leverages it to generate diverse and grounded instructions. and (2) a policy-switching strategy for trajectory rollout. By alternating between learner and expert models, it captures essential error-recovery data often missing in standard imitation learning. Agents trained on our data achieve competitive results across three dynamic mobile agent benchmarks: notably, our fine-tuned Qwen2.5-VL and Qwen3-VL reach 51.7% and 64.7% on AndroidWorld, far surpassing existing open-data approaches. Furthermore, we conduct transparent analyses on the overlap between our synthetic instructions and benchmark test sets, and verify that performance gains stem from broad functionality coverage rather than benchmark overfitting. We release data and code at https://njucckevin.github.io/openmobile/ to bridge the data gap and facilitate broader mobile agent research.
Kanzhi Cheng, Zehao Li, Zheng Ma +11
1Nanjing University · 4Shanghai AI Laboratory · 2SenseTime +3
We present MobileGym, a browser-hosted, lightweight, fully controllable environment for everyday mobile use, targeting interaction fidelity without replicating proprietary backends. It enables two capabilities previously out of reach for everyday apps: verifiable outcome signals through deterministic state-based judging over structured JSON state, and scalable online RL through low-cost parallel rollouts. The full environment state is captured, configured, forked, and compared as structured JSON, and a single server can host hundreds of parallel instances, with about 400 MB memory per instance and about 3 s cold start. A layered state model and a declarative task-definition framework keep state programmability and task creation practical at scale, and a single programmatic judging mechanism delivers both deterministic evaluation verdicts and dense RL rewards. The accompanying MobileGym-Bench provides 416 parameterized task templates, including 256 test and 160 train templates, over 28 apps, with deterministic judges and a structured AnswerSheet protocol that avoids free-text matching failures. In a Sim-to-Real case study, GRPO on Qwen3-VL-4B-Instruct gains +12.8 percentage points on the 256-task test set, and on a 59-task real-device signal subset, real-device execution retains 95.1% of the simulation-side training gain. Project page: https://mobilegym.github.io.
Dingbang Wu, Rui Hao, Haiyang Wang +8
Institute of Automation, Chinese Academy of Sciences · Peking University · The Chinese University of Hong Kong
Online reinforcement learning (RL) with verifiable rewards has driven rapid progress in mathematics, coding, and tool use, but scaling this success to GUI agents remains challenging due to the high cost of interactive environments and the scarcity of realistic training tasks. We introduce MAI-Gym, an open all-in-one platform that unifies GUI agent evaluation, data collection, and online RL training across mobile, computer, and web environments. Through a shared rollout pipeline, MAI-Gym ensures consistent agent behavior across evaluation and training, supports ten established benchmarks, and scales to 20,000 parallel environment instances. Beyond scalable infrastructure, MAI-Gym enables automated synthesis of diverse GUI applications and challenging tasks, including resettable mocked apps and coherent cross-app worlds with execution-validated task verifiers. Using MAI-Gym, we train MAI-Gym-9B, a unified 9B agent for mobile and computer use, achieving state-of-the-art performance among open GUI agents under 10B parameters: 78.4% on MobileWorld and 76.2% on OSWorld-Verified. Importantly, RL in a mocked train-ticketing app transfers to its real-world counterpart, improving task success on a phys- ical phone from 28.8% to 48.1%. We release the framework, synthesis skills, mocked apps, 2,000 verified tasks, and the model checkpoint to support repro- ducible research and further advances in scalable GUI agent training.