Qt AI Assistant 0.92 Released – Introducing the Code Review Agent (original) (raw)
June 12, 2025 by Peter Schneider | Comments
Qt AI Assistant v0.92 introduces the first agentic AI functionality and supports two additional Large Language Models.
Code Review Agent
Do you want to check your code for common issues before sharing it for review with your peers? Or do you want to have the code generated by one LLM verified by another LLM for the ultimate vibe-coding experience? (Note: Vibe-coding is not recommended for production-quality UI development by the author.)
Using the Code Review Agent will improve the quality of your QML code in seconds, especially if you are new to Qt application development or you need a fresh look at your code. The Code Review Agent reviews your code, especially for syntax, structural, and other common issues and provides a suggestion of how to improve the code.
The Code Review Agent can use a secondary LLM to review code in the code editor, generated by the primary LLM or written by a real developer. Furthermore, based on the context, the Code Review Agent decides whether to use the QML linter to better understand possible issues with the code to be reviewed.
The developer can configure a secondary LLM for code reviews in the AI Assistants settings. If no secondary LLM is chosen, then the primary LLM will perform the code review for prompts.
The QML linter will process QML code to identify issues, which are then shared with the review LLM as context for the review. Code reviews focus on fixing errors and structural issues in the code. Developers activate the agent with the /review command.
Codestral Support for QML and C++ Code Completion
For our customers who love European models, we added support for the latest version of Codestral (a code generation model by French company Mistral) as an option for code completion. The performance on the QML100FIM Benchmark is 69%, making it a good option for customers who would like to use a cloud-hosted LLM option.
Table: QML100FIM Code Completion Performance - June 2025
OpenAI GPT-4.1 Mini Support for Prompts
Looking for a cost-efficient cloud-hosted option for prompt-based coding assistance, we identified that GPT-4.1 mini from OpenAI does an equally good (or even better) job than GPT4o or o1. With this release, we are adding the GPT-4.1 mini LLM as an option for prompts in the Qt AI Assistant. OpenAI GPT-4.1 mini scores an acceptable performance of 53% on the QML100 Benchmark. However, the model is not great at creating Qt Quick Test cases or code reviewing, similar to its other OpenAI siblings.
Due to the below 50% performance in code completion, we decided not to add it as an option for code completion.
QML100 Prompt-based Coding Performance - June 2025
How to Upgrade
If you already use the Qt AI Assistant, click the Update button in the Extensions view (remember that you need at least Qt Creator 16.0.1).
If you are new to the Qt AI Assistant, you need to activate the use of external repositories in the extension view to fetch new extensions, such as the Qt AI Assistant. The very first loading of the Qt AI Assistant may take quite some time (over 60 seconds) due to internal file handling issues. We'll fix it in the near future.
Meanwhile… the following smaller enhancements have been made:
- Windows on ARM support
- Llama 3.3 support for prompts changed from fine-tuned to base model