Qwen-AgentWorld
Qwen-AgentWorld is an open language world model for simulating agent environments through reasoning and next-observation prediction. The 35B-A3B checkpoint covers MCP, search, terminal, software engineering, Android, web, and operating-system interactions using language representations. Continued pretraining, supervised training, and reinforcement learning on real interaction trajectories target faithful environment simulation. The technical report appeared June 23, followed by the June 24 weight announcement. This is a distinct language-model specialization rather than a tool-execution application.
2026-06-23
35B total, 3B activated
Hybrid-attention Mixture of Experts language model
Apache-2.0
Specifications
- Parameters
- 35B total, 3B activated
- Architecture
- Hybrid-attention Mixture of Experts language model
- License
- Apache-2.0
- Context Window
- 262,144 tokens
- Type
- text
- Modalities
- text
Benchmark Scores
Advanced Specifications
- Model Family
- Qwen
- Finetuned From
- Qwen3.5-35B-A3B-Base
- API Access
- Not Available
- Chat Interface
- Not Available
Capabilities & Limitations
- Capabilities
- environment simulationnext-state predictionchain-of-thoughtlanguage world modeling
- Known Limitations
- The checkpoint contains language-model weights only, despite visual component definitions in its configuration.Simulated observations are model predictions rather than verified real execution.
- Notable Use Cases
- agent training environmentssimulation researchagent evaluation