Qwen on June 24 unveiled AgentWorld, a native language world model designed to simulate interactions among AI agents across seven domains.
The company also released AgentWorldBench, a benchmark covering the same seven domains for evaluating language world models and agent capabilities.
Both AgentWorld and AgentWorldBench are available through Hugging Face and ModelScope, Qwen said.