Overview
Qwen-RobotNav is a scalable vision-language-navigation model for mobile robots and agentic navigation systems. It is part of the Qwen-Robot Suite and is built on Qwen3-VL. Qwen’s research listing gives its release date as June 15, 2026. (qwen.ai)
Capabilities and uses
- Supports instruction following, point-goal and object-goal navigation, target tracking, autonomous driving, and navigation for embodied question answering through a shared waypoint-prediction interface.
- Uses a four-layer action head to predict eight waypoints, each represented by
(x, y, theta). An upper-level planner can select task modes and configure visual token budget, temporal decay, camera weights, and frame sampling. - Was trained on 15.6 million samples. The official repository reports zero-shot deployment on a Unitree Go2 robot in previously unseen environments; its stated 196 ms latency and 5.1 Hz rate apply to the demonstrated Jetson Thor setup. (github.com)
Limitations and availability
Qwen’s official repository states that there is currently no plan to release Qwen-RobotNav model weights. The reported real-world demonstrations and benchmark results do not establish reliability or safety in every environment. A public API model ID was not verified. (github.com)
Pricing
As of November 16, 2026, Qwen-RobotNav-specific input and output charges could not be verified on either Alibaba Cloud’s mainland China pricing page in CNY/RMB or its international pricing page in USD. No price or billing unit is assumed. (help.aliyun.com)
Official sources
- Qwen-Robot Suite announcement — the supplied page was opened, but its article text was not readable through the browsing tool.
- Qwen research listing — lists the model and release date. (qwen.ai)
- Official Qwen-RobotNav repository — model details, evaluations, and weight availability. (github.com)
- Alibaba Cloud Suite overview. (alibabacloud.com)
- Pricing checked November 16, 2026: mainland China pricing and international pricing. (help.aliyun.com)