[AutoBe] Built full-level backend applications with "qwen-next-80b-a3b" model

The AutoBE team has successfully tested the qwen3-next-80b-a3b-instruct model to generate three complete backend applications: To Do List, Reddit Community, and Economic Discussion Board.

Project Implementations

We created working examples across three model variants:

• qwen3-next-80b-a3b-instruct
• openai/gpt-4.1-mini
• openai/gpt-4.1

Each model generated versions of:

• To Do List applications
• Reddit Community platforms
• Economic Discussion Boards (BBS)
• E-Commerce systems

Key Findings

While qwen3-next-80b-a3b-instruct failed during the realize phase, this stemmed from compiler development rather than model limitations. AutoBE addresses this through AI-friendly compilers providing error feedback to agents.

Compared to OpenAI models, the Qwen variant produced fewer documents, API operations, and schemas but demonstrated superior cost efficiency. We observed limited E2E test generation — the Reddit project contained only 9 tests covering 60 API operations.

Assessment

We consider these implementations successful despite remaining compilation issues. There are areas needing improvement, but Qwen’s economic advantages are significant, particularly relevant for an open-source project prioritizing community accessibility.

• GitHub Repository 
• Documentation