JetBrains, the company famous for developing IntelliJ IDE and Kotlin language, recently launched its first AI coding assistant Junie. The new tool is designed to help developers handle day-to-day development tasks and understand the context of existing projects to support the creation of new applications and feature expansion.
Junie successfully solved 53.6% of 500 common developer tasks on the highly regarded SWE-Bench Verified benchmark. Although this result is not the highest at present (the current best model scores more than 60%), JetBrains believes that Junie's performance is "very promising." It is worth noting that the currently leading Weights & Biases “Programmer O1crosscheck5” model scored 64.6%.
Junie's unique advantage is its deep integration with JetBrains IDE. JetBrains emphasizes that even if Junie assists developers in completing their work, human developers are always in control. Not only will this tool speed up development, it will also help improve code quality. By combining the power of JetBrains IDE with Large Language Models (LLMs), Junie can generate code, run checks, write tests, and verify its validity.
"AI-generated code can be just as flawed as code written by developers," JetBrains said in the announcement. "Ultimately, Junie will not only speed up development, but also raise the bar for code quality."
Junie is currently only available through a waitlisted early access program and only supports IntelliJ IDEA Ultimate and PyCharm Professional IDE on Linux and Mac systems, with support for WebStorm coming soon.
The launch of this AI coding assistant marks another important move by JetBrains in the field of AI-assisted development. Junie's deep integration and emphasis on human developer control may set it apart in the highly competitive market for AI coding assistants.
