QwenLM/Qwen-AgentWorld

QwenLM/Qwen-AgentWorld

Description: Qwen-AgentWorld: Language World Models for General Agents

Language: Python

License: Apache-2.0

Stars: 70

Forks: 1

Open issues: 0

Created: 2026-06-22T13:48:37Z

Pushed: 2026-06-24T02:10:42Z

Default branch: main

Fork: no

Archived: no

README:

Qwen-AgentWorld

📑 Technical Report | 📖 Blog | 🤗 Hugging Face | 🤖 ModelScope | 🖥️ Demo

Welcome to the GitHub repository of Qwen-AgentWorld. Here, you can find official information about Qwen-AgentWorld, post your questions (Issues), and share your ideas with the community (Discussions).

News

• 2026-06-24: We release Qwen-AgentWorld-35B-A3B and AgentWorldBench. Read more on our blog and the technical report.

Open-Source Release

We open-source Qwen-AgentWorld-35B-A3B (model weights) and AgentWorldBench (evaluation benchmark):

| Release | Description | |---------|-------------| | Qwen-AgentWorld-35B-A3B | Language world model (MoE, 35B total / 3B active, 256K context) | | AgentWorldBench | Evaluation benchmark across 7 domains |

The official weights and data are released on:

• 🤗 HuggingFace: Download automatically via model ID, e.g., Qwen/Qwen-AgentWorld-35B-A3B. You can also download model files manually using huggingface download or git clone. Please follow the instructions on the model page.
• 🤖 ModelScope: For users unable to access Hugging Face Hub. For supported frameworks, you can download from ModelScope by setting environment variables, such as SGLANG_USE_MODELSCOPE=true or VLLM_USE_MODELSCOPE=true.

Introduction

Qwen-AgentWorld is a native language world model that simulates agentic environments via long chain-of-thought reasoning across seven unified domains: MCP, Search, Terminal, SWE, Android, Web, and OS. It is trained through a three-stage pipeline -- CPT injects environment knowledge, SFT activates next-state-prediction reasoning, RL sharpens simulation fidelity -- on more than 10M real-world interaction trajectories. Unlike prior approaches that treat world modeling as a post-hoc add-on, Qwen-AgentWorld is a native world model: environment modeling is the training objective from the CPT stage onward.

Key features:

• Seven Unified Domains. The first language world model to cover seven agent interaction domains within a single model.
• Native World Model. Environment modeling from CPT onward, not post-hoc adaptation.
• Generalizable, Scalable & Controllable Simulator. Zero-shot generalization to OOD environments (e.g., Claw Agent); controllable perturbations and fictional-world construction surpass real-environment training.
• Agent Foundation Model. LWM RL warm-up on single-turn, non-agentic trajectories transfers to multi-turn, tool-calling agentic tasks across seven benchmarks, including three entirely out-of-domain.

Performance

Five-dimensional rubric mean (↑) per domain, normalized to 0--100 scale.

| Model | MCP | Search | Term. | SWE | Android | Web | OS | Overall | |:------|:---:|:------:|:-----:|:---:|:-------:|:---:|:--:|:-----------:| | GPT-5.4 | 70.10 | 37.26 | 53.69 | 66.29 | 60.00 | 51.80 | 68.58 | 58.25 | | Claude Opus 4.8 | 54.93 | 35.14 | 59.18 | 64.10 | 61.50 | 54.66 | 66.62 | 56.59 | | Claude Opus 4.6 | 69.90 | 29.30 | 57.51 | 64.55 | 61.74 | 51.42 | 70.20 | 57.80 | | Gemini 3.1 Pro | 59.07 | 30.21 | 52.47 | 59.07 | 61.40 | 52.83 | 66.92 | 54.57 | | Claude Sonnet 4.6 | 70.00 | 28.79 | 56.98 | 64.52 | 58.03 | 50.78 | 63.17 | 56.04 | | DeepSeek-V4-Pro | 63.27 | 27.61 | 51.26 | 59.44 | 55.17 | 50.32 | 63.70 | 52.97 | | GLM-5.1 | 67.60 | 22.46 | 47.32 | 52.07 | 59.10 | 51.50 | 59.13 | 51.31 | | Kimi K2.6 | 65.23 | 27.48 | 52.54 | 58.77 | 58.93 | 50.20 | 60.80 | 53.42 | | MiniMax-M2.7 | 55.82 | 27.30 | 41.62 | 37.44 | 52.40 | 50.52 | 57.73 | 46.12 | | Qwen3.5-35B-A3B | 57.87 | 25.98 | 46.13 | 47.58 | 53.18 | 47.10 | 56.27 | 47.73 | | Qwen3.5-397B-A17B | 68.31 | 30.81 | 55.30 | 64.44 | 54.90 | 48.55 | 60.85 | 54.74 | | Qwen3.6-Plus | 55.28 | 21.94 | 50.58 | 59.08 | 57.65 | 50.78 | 60.33 | 50.81 | | Qwen-AgentWorld-35B-A3B | 64.79 | 36.69 | 53.96 | 65.63 | 58.17 | 49.55 | 65.92 | 56.39 | | Qwen-AgentWorld-397B-A17B | 68.24 | 37.82 | 57.73 | 68.49 | 60.20 | 50.98 | 67.89 | 58.71 |

Qwen-AgentWorld-397B-A17B achieves the highest overall score (58.71), outperforming all frontier proprietary models including GPT-5.4 (58.25). Qwen-AgentWorld-35B-A3B shows +8.66 improvement over Qwen3.5-35B-A3B without LWM training.

Applications

Generalizable Environment Scaling. Sim RL with Qwen-AgentWorld-397B-A17B on 4k out-of-distribution OpenClaw environments:

| Model | Claw-Eval | QwenClawBench | |:------|:---------:|:-------------:| | Qwen3.5-35B-A3B | 65.4 | 47.9 | | + Sim RL (w/ Qwen3.6-Plus) | 66.7 | 47.8 | | + Sim RL (w/ Qwen-AgentWorld-397B-A17B) | 69.7 | 55.0 | | Δ | +4.3 | +7.1 |

Controllable Simulation: MCP. Environment Adaptation --- Control instructions inject targeted perturbations to expose agent weaknesses:

| Model | Tool Decathlon | MCPMark | |:------|:--------------:|:-------:| | Qwen3.5-35B-A3B-SFT | 32.4 | 21.5 | | + Sim RL (uncontrolled) | 31.5 | 24.6 | | + Sim RL (controlled) | 36.1 | 33.8 | | Δ | +3.7 | +12.3 |

Controllable Simulation: Search. Fictional-World Construction -- agents trained in fully invented, self-consistent worlds generalize to real search tasks:

| Model | WideSearch F1 Item | WideSearch F1 Row | |:------|:------------------:|:-----------------:| | Qwen3.5-35B-A3B-SFT | 34.02 | 13.72 | | + Sim RL (controlled) | 50.31 | 24.21 | | Δ | +16.29 | +10.49 | | | | | | Qwen3.5-397B-A17B-SFT | 70.11 | 45.69 | | + Sim RL (controlled) | 73.98 | 51.74 | | Δ | +3.87 | +6.05 |

Agent Foundation Model. LWM RL warm-up on single-turn, non-agentic trajectories transfers to multi-turn, tool-calling agentic tasks:

| | Terminal-Bench 2.0 | SWE-Bench Verified | SWE-Bench Pro | WideSearch F1 Item | Claw-Eval | QwenClawBench | BFCL v4 | |:---|:---:|:---:|:---:|:---:|:---:|:---:|:---:| | | *In Domain* | | | | *Out of...