LLM-in-Sandbox

🌐 Project Page • 📄 Paper • 🤗 Huggingface

Enabling LLMs to explore within a code sandbox (i.e., a virtual computer) to elicit general agentic intelligence.

▶️ Click to watch the demo video

Features:

• 🌍 General-purpose: works beyond coding—scientific reasoning, long-context understanding, video production, travel planning, and more
• 🐳 Isolated execution environment via Docker containers
• 🔌 Compatible with OpenAI, Anthropic, and self-hosted servers (vLLM, SGLang, etc.)
• 📁 Flexible I/O: mount any input files, export any output files

Table of Contents

• Installation
• Quick Start
• More Examples
• Benchmark and Reproduction
• Citation

Installation

Requirements: Python 3.10+, Docker

pip install llm-in-sandbox

Or install from source:

git clone https://github.com/llm-in-sandbox/llm-in-sandbox.git
cd llm-in-sandbox
pip install -e .

Docker Image

The default Docker image (cdx123/llm-in-sandbox:v0.1) will be automatically pulled when you first run the agent. The first run may take a minute to download the image (~400MB), but subsequent runs will start instantly.

llm-in-sandbox build

Quick Start

LLM-in-Sandbox works with various LLM providers including OpenAI, Anthropic, and self-hosted servers (vLLM, SGLang, etc.).

Option 1: Cloud / API Services

llm-in-sandbox run \
    --query "write a hello world in python" \
    --llm_name "openai/gpt-5" \
    --llm_base_url "http://your-api-server/v1" \
    --api_key "your-api-key"

Option 2: Self-Hosted Models

vllm serve Qwen/Qwen3-Coder-30B-A3B-Instruct \
    --served-model-name qwen3_coder \
    --enable-auto-tool-choice \
    --tool-call-parser qwen3_coder \
    --tensor-parallel-size 8  \
    --enable-prefix-caching

llm-in-sandbox run \
    --query "write a hello world in python" \
    --llm_name qwen3_coder \
    --llm_base_url "http://localhost:8000/v1"  \
    --temperature 0.7

python3 -m sglang.launch_server \
    --model-path "deepseek-ai/DeepSeek-V3.2" \
    --served-model-name "DeepSeek-V3.2" \
    --trust-remote-code \
    --tp-size 8 \
    --tool-call-parser deepseekv32 \
    --reasoning-parser deepseek-v3 \
    --host 0.0.0.0 \
    --port 5678

llm-in-sandbox run \
    --query "write a hello world in python" \
    --llm_name DeepSeek-V3.2 \
    --llm_base_url "http://0.0.0.0:5678/v1" \
    --extra_body '{"chat_template_kwargs": {"thinking": True}}'

Parameters (Common)

Run llm-in-sandbox run --help for all available parameters.

Output

Each run creates a timestamped folder:

output/2026-01-16_14-30-00/
├── files/
│   ├── answer.txt      # Final answer
│   └── hello_world.py  # Output file
└── trajectory.json     # Execution history

More Examples

We provide examples across diverse non-coding domains: scientific reasoning, long-context understanding, instruction following, travel planning, video production, music composition, poster design, and more.

👉 See examples/README.md for the full list.

Benchmark and Reproduction

Reproduce our paper results, evaluate any LLM in the sandbox, or add your own tasks.

👉 See llm_in_sandbox/benchmark/README.md

Contact Us

Daixuan Cheng: daixuancheng6@gmail.com
Shaohan Huang: shaohanh@microsoft.com

Acknowledgment

We learned the design and reused code from R2E-Gym. Thanks for the great work!

Citation

If you find our work helpful, please cite us:

@article{cheng2026llm,
  title={LLM-in-Sandbox Elicits General Agentic Intelligence},
  author={Cheng, Daixuan and Huang, Shaohan and Gu, Yuxian and Song, Huatong and Chen, Guoxin and Dong, Li and Zhao, Wayne Xin and Wen, Ji-Rong and Wei, Furu},
  journal={arXiv preprint arXiv:2601.16206},
  year={2026}
}