Description
Qwen-AgentWorld-35B-A3B is a sophisticated language world model developed to simulate agentic environments across seven distinct domains. These domains include tool calling, search, terminal, software engineering, Android, web, and operating systems. The model is trained through a three-stage pipeline consisting of Continual Pre-Training (CPT), Supervised Fine-Tuning (SFT), and Reinforcement Learning (RL). This approach ensures that environment modeling is integral to its functionality, rather than an afterthought.
Read More
The model is built on the Qwen3.5-35B-A3B-Base and features 35 billion parameters, with 3 billion activated during operation. It employs a complex architecture with 40 layers and a context length of 262,144 tokens, allowing for extensive multi-turn environment simulations. The model's ability to generalize to out-of-domain environments and its controllable perturbations make it a versatile tool for various applications.
Qwen-AgentWorld-35B-A3B is compatible with popular inference frameworks such as Hugging Face Transformers, vLLM, and SGLang. It supports zero-shot generalization and can be deployed via APIs, making it accessible for developers looking to integrate advanced simulation capabilities into their applications. The model's performance is evaluated using the AgentWorldBench, where it scores across five dimensions: format, factuality, consistency, realism, and quality.
This model is particularly useful for developers and researchers in fields requiring complex environment simulations, such as AI research, software development, and interactive system design. Its open-source nature and compatibility with existing frameworks enhance its appeal to a broad audience.
- CategoryAI Models & LLMs
- Pricing ModelOpen source
- DeploymentCloud
- PlatformsWeb
Qwen-AgentWorld-35B-A3B Highlights
Seven Unified Domains
Native World Model
Generalizable Simulator
35 Billion Parameters
Context Length: 262,144 tokens
Open Source
API Available
Reinforcement Learning Training
Getting Started with Qwen-AgentWorld-35B-A3B
Access model page: Visit the Hugging Face repository
Authenticate: Ensure access to necessary resources
Configure: Set up model parameters and environment
Send prompts: Use domain-specific system prompts
Fine-tune: Adjust model settings for specific tasks
Qwen-AgentWorld-35B-A3B's Use Cases
- Tool Calling
- Search Simulation
- Terminal Operations
- Software Engineering
- Android Interaction
