Software development company JetBrains announced that it has open-sourced Mellum, an AI model designed from the ground up to enable cloud-based code completion. Available on Hugging Face, Mellum offers researchers, educators, and advanced teams the opportunity to explore its inner workings and understand how a purpose-built model functions in practice.
Unlike typical large language models (LLMs), Mellum is specifically optimized for coding. The company focused on creating a smaller and more efficient model that minimizes latency, providing developers with almost instant code suggestions. This makes Mellum more effective than many third-party models currently on the market.
Mellum supports code completion for a range of programming languages, including Java, Kotlin, Python, Go, PHP, C, C++, C#, JavaScript, TypeScript, CSS, HTML, Rust, and Ruby.
With 4 billion parameters, Mellum is a multilingual model specifically that has been benchmarked across various datasets and languages, and extensive human evaluations have been conducted within JetBrains’ integrated development environments (IDEs). The current version of Mellum available on Hugging Face is intended for use by AI/ML researchers—particularly those examining AI’s role in software development, model benchmarking, or interpretability. It also serves as a foundation for AI/ML engineers and educators who are interested in learning how to build, fine-tune, or adapt domain-specific language models, or who are looking to support educational programs centered around LLM architecture and specialization.
Mellum is now open source on @huggingface!
It’s a focal model that is small, efficient, and made for one thing: code completion.
Trained from scratch by JetBrains.
First in a growing family of dev-focused LLMs.
https://t.co/NPrKQunIeS pic.twitter.com/EXM8mhCYXp
— JetBrains (@jetbrains) April 30, 2025
JetBrains’ Mellum Integration Powers AI Assistant With Enhanced Code Completion Performance
Since Mellum has been integrated into AI Assistant to power code completion, it has produced notable improvements. Completion latency has been reduced to a third of its previous time, indicating a substantial decrease in the time needed to complete tasks. The acceptance rate of completion suggestions driven by the Mellum AI model stands at around 40%, which is considered a solid benchmark in the industry. Additionally, the cancel rate has decreased by three to four times compared to the earlier version of AI Assistant. Furthermore, the total number of completion suggestions displayed has more than doubled when compared to the previous version.
JetBrains is known for its specialization in creating intelligent tools for software developers and teams. It is particularly recognized for its IDEs that support a wide range of programming languages. One of JetBrains’ major contributions to the software development field is Kotlin, a statically typed programming language designed to run on the Java Virtual Machine (JVM).
The post JetBrains Open-Sources Mellum AI Model For Cloud-Based Code Completion, Now Available On Hugging Face appeared first on Metaverse Post.