In the rapidly evolving world of technology and development, the intersection of AI and coding is becoming increasingly significant. For those in the cryptocurrency and blockchain space, understanding the tools that shape the future of software development is key. A notable step in this direction comes from JetBrains, a company well-known for its comprehensive suite of developer tools. They have recently unveiled something particularly interesting: JetBrains Mellum , their first truly ‘open’ AI model designed specifically for coding. Introducing JetBrains Mellum: A New AI Coding Model JetBrains, the company behind popular integrated development environments (IDEs) like IntelliJ IDEA, PyCharm, and others, has made a significant move by releasing its own AI coding model , Mellum. This model, which was previously integrated into JetBrains’ commercial tools, is now openly available on the AI development platform Hugging Face. This move signals a commitment to contributing to the broader AI research and development community. Key details about Mellum: It’s designed primarily for code completion, helping developers write code faster and more efficiently by suggesting snippets based on context. Mellum is a relatively compact model, weighing in at 4 billion parameters. It was trained on a substantial dataset of over 4 trillion tokens. The training involved a cluster of 256 H200 Nvidia GPUs and took around 20 days. The model is released under the permissive Apache 2.0 license. Understanding the scale: Parameters roughly indicate a model’s complexity and problem-solving capacity, while tokens represent the raw data it processed during training. JetBrains notes that a million tokens correspond to about 30,000 lines of code, giving a sense of the vast amount of information Mellum learned from. Why an Open Source AI Model Matters The decision by JetBrains to release Mellum as an open source AI model on Hugging Face is noteworthy. While many powerful AI models remain proprietary, making Mellum openly available fosters transparency, collaboration, and further innovation within the developer community. An open model allows researchers, developers, and enthusiasts to: Inspect the model’s architecture and training data (where permissible). Experiment with its capabilities. Fine-tune it for specific tasks or programming languages. Integrate it into various tools and workflows beyond JetBrains’ own ecosystem. JetBrains explicitly states that Mellum is designed for integration into professional developer tooling, AI-powered coding assistants, and research on code understanding and generation. Its open nature also makes it suitable for educational purposes and fine-tuning experiments. Focus on Code Completion AI Unlike larger, more general-purpose AI models that can write essays or generate images, Mellum has a specific focus: code completion AI . This specialization allows the model to be highly optimized for the task of predicting and suggesting code snippets within an IDE or code editor. This is a core function of modern development environments and can significantly boost developer productivity. However, JetBrains is clear that the base Mellum model requires fine-tuning before it can be effectively used. They have provided some fine-tuned examples for Python, but these are intended for evaluation and estimation of potential capabilities, not for production deployment. This highlights that while the core model is open, practical application still requires effort to adapt it to specific needs and contexts. Implications for AI Development Tools and Security The release of Mellum contributes to the growing landscape of AI development tools . As AI becomes more integrated into the software development lifecycle, tools like Mellum will become essential for developers. They promise increased speed and efficiency, but also introduce new considerations, particularly regarding security. AI-generated code, while helpful, is not immune to vulnerabilities. A survey from late 2023 by Synk indicated that over 50% of organizations frequently encounter security issues with AI-produced code. JetBrains acknowledges this, noting that Mellum’s suggestions may reflect biases present in the public codebases it was trained on and won’t necessarily be secure or free of vulnerabilities. Developers must still apply rigorous testing and security practices when using AI-assisted coding. Despite the challenges, JetBrains sees this release as a starting point. Their goal isn’t necessarily generality but focus and contribution to the community. They hope Mellum will spark meaningful experiments and collaborations. In conclusion, JetBrains’ release of Mellum as an open AI coding model is a significant development for the software industry. By focusing on code completion AI and making the model available on Hugging Face, they are contributing to the ecosystem of AI development tools and fostering open research. While challenges like fine-tuning and security require attention, JetBrains Mellum represents a valuable resource for developers and researchers exploring the future of AI-assisted coding. To learn more about the latest AI development tools trends, explore our article on key developments shaping AI features.
