Automated code generation and testing with Qwen 2.5 Max – A showcase

Automated code generation and testing with Qwen 2.5 Max – A showcase

The world of artificial intelligence is currently undergoing a tremendous transformation. With the release of powerful models like Alibaba's Qwen 2.5 Max, new possibilities are opening up in software development. One particularly exciting application area is the combination of automated code generation and testing . In this article, we demonstrate how Qwen 2.5 Max can be used to generate code and simultaneously test it according to ISTQB criteria. We also examine the advantages and disadvantages of this approach and discuss potential challenges.

Qwen 2.5 Max – An Overview

Qwen 2.5 Max is a powerful AI model from Alibaba that has achieved a remarkable performance of 38.7% . It is specifically optimized for code processing and generation and offers improved contextual understanding. This makes the model suitable not only for code generation but also for automated test creation.

What is LiveCodeBench?

LiveCodeBench is a benchmarking tool for evaluating AI models in terms of their ability to generate and understand code. A higher percentage, such as 38.7% , means that the model performs above average in a range of standardized test tasks and produces high-quality code.

You can find more information about LiveCodeBench here: LiveCodeBench website

Showcase: Writing and testing code with Qwen 2.5 Max

We will have Qwen 2.5 Max generate a simple function for primality checking, as well as the unit tests, which should be ISTQB compliant.

Prompt: We want to implement a TypeScript function that checks whether a given number is prime. Additionally, we should write unit tests using Jest that meet the ISTQB criteria.

Function generated by Qwen 2.5 Max and associated ISTQB-compliant unit tests using Jest

Run tests

To run the function and tests created by Qwen 2.5 Max, we create a project in Microsoft Visual Studio Code and copy the code for checking prime numbers, as well as the generated unit tests, into it. Then we install Jest and run the tests.

Test result

As the example below shows, the unit tests generated by Qwen 2.5 Max could be successfully executed without problems in a development environment such as Microsoft Visual Studio Code.

Analysis of the results

Advantages of AI-powered code generation and verification

✅ Efficiency: Code and tests can be generated within seconds.
✅ Automated test cases: AI ensures that the tests cover various scenarios. In our case, test cases were generated according to ISTQB criteria (boundary analyses, good case/bad case, performance tests, etc.).
✅ Reduction of human error: AI minimizes typos or logical errors in the code.

Disadvantages and challenges

⚠️ Quality assurance: The generated code must be manually reviewed to ensure it is correct and performs well. Without thorough analysis, faulty or inefficient code can result.

⚠️ Complexity: AI struggles with complex software structures. End-to-end (E2E) tests, in particular, which require complete system validation, can only be partially automated because they often rely on integration and user interaction.

⚠️ Dependence on training data: The model is based on existing data and generates code according to learned patterns. This can limit innovation and customized solutions. Developers often have to adapt the generated code afterward to meet specific requirements.

Conclusion

Qwen 2.5 Max demonstrates impressive capabilities in automated code generation and testing. The AI ​​can provide valuable support, especially for standard applications and unit tests. However, human review is essentialto ensure the code is of high quality and meets all requirements. In the future, such models could play an even greater role in software development and significantly accelerate the development process.

Targeted use of AI in software development

The world of artificial intelligence is developing rapidly, opening up new possibilities for code generation and testing. But which AI tools are best suited to your specific requirements?

At Infometis, we specialize in various AI-powered solutions and are happy to provide you with personalized advice. If you would like to learn more about how you can effectively integrate AI into your development processes, feel free to schedule a no-obligation appointment with us at any time.

👉 Get in touch now!