JetBrains has open-sourced its Mellum large language model (LLM), which was purpose-built for code completion, with the aim to grow Mellum into a family of models specialized for different coding tasks.
The Mellum base model was open-sourced and made available on Hugging Face on April 30. By releasing Mellum on Hugging Face, JetBrains is offering researchers, educators, and advanced teams the opportunity to explore how a purpose-built model works under the hood, JetBrains said. The company trained Mellum to power cloud-based code completion in JetBrains IDEs. However, Mellum is not aimed at the average developer, but is best suited for AI and machine learning researchers, engineers, and educators who want to explore, fine-tune, or teach domain-specific LLMs in the context of software development.
Mellum is being open-sourced because JetBrains believes in transparency, collaboration, and the power of shared progress, JetBrains said. Open source has driven big leaps in technology in areas such as Linux, Git, Node.js, and Docker, and now open source LLMs are outperforming some proprietary industry leaders, according to JetBrains. Released to the public last year, Mellum supports code completion for languages including Java, Kotlin, Python, Go, PHP, C, C++, C#, JavaScript, TypeScript, CSS, HTML, Rust, and Ruby.
JetBrains open-sources Mellum LLM | InfoWorldRead More
