JetBrains, the renowned developer of popular app development tools, has taken a significant step forward in the realm of artificial intelligence (AI) by releasing its first open AI model for coding, dubbed Mellum. This move marks a major milestone in the company's efforts to democratize access to AI-powered coding tools, making it possible for developers to integrate the technology into their workflows.

Mellum, initially released as a closed model for JetBrains' software development suites last year, is now openly available on the Hugging Face platform. This AI model is specifically designed for code completion, capable of generating code snippets based on surrounding context. With an impressive 4 billion parameters and trained on over 4 trillion tokens, Mellum is poised to revolutionize the way developers work.

To put Mellum's capabilities into perspective, a million tokens roughly correspond to 30,000 lines of code. The model's parameters, which roughly correspond to its problem-solving skills, make it an extremely powerful tool for developers. JetBrains has made Mellum available under the Apache 2.0 license, ensuring that it can be freely used, modified, and distributed.

The training process for Mellum was no small feat, requiring around 20 days on a cluster of 256 H200 Nvidia GPUs. The model was trained on a diverse collection of datasets, including permissively licensed code from GitHub and English-language Wikipedia articles. This diverse training set enables Mellum to generate code that is not only functional but also follows best practices and conventions.

While Mellum holds immense potential, it does require fine-tuning before it can be used in production environments. JetBrains has provided a few fine-tuned models for Python, but these are intended for estimation purposes only. Developers will need to invest time and effort into fine-tuning Mellum for their specific use cases.

The release of Mellum comes at a time when AI-generated code is becoming increasingly prevalent in software development. However, this trend also introduces new security challenges. According to a survey by developer security platform Synk, more than 50% of organizations encounter security issues with AI-produced code sometimes or frequently. JetBrains acknowledges that Mellum may "reflect biases present in public codebases" and that its code suggestions may not always be "secure or free of vulnerabilities."

Despite these challenges, JetBrains remains optimistic about the potential of Mellum. The company sees this release as "just the beginning" and hopes that it will spark meaningful experiments, contributions, and collaborations within the developer community. With Mellum now openly available, the possibilities for innovation and advancement in AI-powered coding tools are vast and exciting.

As the AI landscape continues to evolve, the release of Mellum is a significant step forward in making AI-powered coding tools more accessible and usable. As developers begin to explore the capabilities of Mellum, it will be interesting to see the impact it has on the software development industry as a whole.