--- ## Highlights - **#1 Open-Source Model on OSWorld**: Achieves **56.7%** task completion rate, **#1 among all open-source models** - **Significant Improvements**: +11.7% over OpenCUA-72B (45.0%→56.7%), +15.1% over Qwen3-VL thinking (41.6%→56.7%), with fewer parameters and half the steps - ️ **End-to-End Multi-Turn Automation**: Operates Chrome, Excel, PowerPoint, VSCode and more through screenshots and natural language instructions - **Novel Training Method**: Our data synthesis and training approach consistently improves Computer Use capability across multiple open-source VLMs without degrading general performance --- ## Performance Comparison | Rank | Model | Open/Closed | Type | Max Steps | Score | |------|-------|-------------|------|-----------|-------| | 1 | Claude-sonnet-4-5 | Closed | General | 100 | 62.9% | | 2 | Seed-1.8 | Closed | General | 100 | 61.9% | | 3 | Claude-sonnet-4-5 | Closed | General | 50 | 58.1% | | **4** | **EvoCUA-20260105 (Ours)** | ** Open** | **General** | **50** | **56.7% ** | | 5 | DeepMiner-Mano-72B | Closed | Specialized | 100 | 53.9% | | 6 | UI-TARS-2-2509 | Closed | General | 100 | 53.1% | | 7 | EvoCUA (Previous Version) | Closed | General | 50 | 50.3% | | 8 | OpenCUA-72B | Open | Specialized | 100 | 45.0% | | ... | ... | ... | ... | ... | ... | | 13 | Qwen3-VL-Flash | Closed | General | 100 | 41.6% | > EvoCUA is **#1 among all open-source models**, achieving competitive results with only **50 steps**. Human-level performance remains significantly higher, indicating substantial room for improvement. --- ## Quick Start ### Installation Python 3.12 is recommended. \`\`\`bash git clone https://github.com/meituan/EvoCUA.git cd EvoCUA python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt \`\`\` ### Model Download & Deployment EvoCUA requires downloading the model weights from HuggingFace and deploying with **vLLM** as an OpenAI-compatible inference server. Recommended versions: - torch: 2.8.0+cu126 - transformers: 4.57.3 - vllm: 0.11.0 \`\`\`bash # 1) Download model weights huggingface-cli download meituan/EvoCUA-32B-20260105 \ --local-dir /path/to/EvoCUA-32B \ --local-dir-use-symlinks False # 2) Launch vLLM serving (recommend separate environment) vllm serve /path/to/EvoCUA-32B \ --served-model-name EvoCUA \ --host 0.0.0.0 \ --port 8080 \ --tensor-parallel-size 2 # 3) Set environment variables # Environment variables can be configured in .env file (see env.template for reference): cp env.template .env # Edit .env with your configurations, e.g., export OPENAI_API_KEY="dummy" export OPENAI_BASE_URL="http://127.0.0.1:8080/v1" \`\`\` ### Run Evaluation on OSWorld \`\`\`bash python3 run_multienv_evocua.py \ --headless \ --provider_name aws \ --observation_type screenshot \ --model EvoCUA-S2 \ --result_dir ./evocua_results \ --test_all_meta_path evaluation_examples/test_nogdrive.json \ --max_steps 50 \ --num_envs 30 \ --temperature 0.01 \ --max_history_turns 4 \ --coordinate_type relative \ --resize_factor 32 \ --prompt_style S2 \`\`\` --- ## Project Structure \`\`\` EvoCUA/ ├── run_multienv_evocua.py # Main entry point (multi-env parallel evaluation) ├── lib_run_single.py # Single task rollout logic (trajectory, screenshots, recording, scoring) ├── lib_results_logger.py # Real-time result aggregation to results.json ├── desktop_env/ # OSWorld environment implementation │ ├── providers/ # VM providers (AWS/VMware/Docker/etc.) │ ├── controllers/ # Environment controllers │ └── evaluators/ # Task evaluators ├── mm_agents/ │ └── evocua/ # EvoCUA agent (prompts, parsing, action generation) └── evaluation_examples/ # OSWorld task configurations \`\`\` --- ## About OSWorld [OSWorld](https://os-world.github.io/) is the most influential benchmark in the Computer Use Agent domain. It is adopted by leading AI organizations including **OpenAI, Anthropic, ByteDance Seed, Moonshot AI, Zhipu AI, Step**, and more. OSWorld evaluates agents' ability to complete real-world computer tasks through multi-turn interactions with actual desktop environments. --- ## Resources - **Model Weights**: [meituan/EvoCUA-32B-20260105](https://huggingface.co/meituan/EvoCUA-32B-20260105) - **OSWorld Benchmark**: [os-world.github.io](https://os-world.github.io/) - **Technical Report**: Coming Soon - **More Model Sizes**: Models of various sizes are on the way and will be open-sourced soon! --- ## Acknowledgements We sincerely thank the open-source community for their outstanding contributions to the Computer Use Agent field. We are grateful to **Xinyuan Wang** ([OpenCUA](https://github.com/xlang-ai/OpenCUA)) and **Tianbao Xie** ([OSWorld](https://github.com/xlang-ai/OSWorld)) for their insightful discussions, valuable feedback on evaluation, and continuous support throughout this project. Their pioneering work has greatly inspired and advanced our research. We are committed to giving back to the community and will continue to open-source our research to advance the field. --- ## Citation If you find EvoCUA useful in your research, please consider citing: \`\`\`bibtex @misc\{evocua2026, title=\{EvoCUA: Evolving Computer Use Agent\}, author=\{Chong Peng* and Taofeng Xue*\}, year=\{2026\}, url=\{https://github.com/meituan/EvoCUA\}, note=\{* Equal contribution\} \} \`\`\` --- ## License This project is licensed under the Apache 2.0 License - see the [LICENSE](./LICENSE) file for details. ---