Presented by Febria
📅 Submission Deadline: July 9, 2026 at 2:00 PM PT → Submit on Devpost
🔄 Last Updated: June 30, 2026 — Updated dynamically as new questions arise
💬 Still have a question? Post it in the Devpost Discussions tab or ask inside the official Qwen Cloud Discord channel. The Alibaba and Devpost team monitors both portals daily.
One application, one track. Your project can only be submitted to a single track. If your application features a mixture of cross-track capabilities, choose the track matching its core technical breakthrough:
No. Evaluation tracks such as Autopilot Agent or AI Showrunner do not require active generation of profit or existing market revenue. "Problem Value & Impact" implies that you have thoroughly mapped out the commercial scaling model, productization path, or open-source community adoption potential to address an authentic business pain point. Show your strategic logic and engineering execution, not an active profit-and-loss sheet.
If you are unsure where your AI application fits, choose from the five specialized focus areas:
| Track | Core Objectives & Focus Areas |
|---|---|
| MemoryAgent | Building agents utilizing persistent, structured memory architectures that dynamically adapt and recall deep context across separate execution sessions. |
| AI Showrunner | Designing automated, generative content production pipelines capable of outputting video assets, audio layers, descriptive scripts, or multi-modal arrays. |
| Agent Society | Deploying complex multi-agent ecosystems that collaborate, negotiate, split tasks, or compete within a shared environmental simulation. |
| Autopilot Agent | Developing end-to-end autonomous software networks capable of handling real-world operations and macro business tasks with zero human oversight. |
| EdgeAgent | Engineering hyper-optimized agents specialized for on-device inferencing, prioritizing ultra-low latency within highly constrained hardware profiles. |
Your application must directly utilize Qwen Cloud hosted models and API frameworks. The model marketplace provides a suite of multi-modal, reasoning, and media generation foundations:
qwen3.7-max (optimized for hyper-complex logic patterns) and qwen3.7-plus (balanced speed and operational performance).