JetBrains releases Mellum 2.1, an open-source model for agentic coding

JetBrains has released Mellum 2.1, a new open-source model that the company positions around stronger agentic coding, with support for local deployment and efficient inference.

That framing points to letting AI take a deeper part in code work. Agentic coding means the model does more than autocomplete code; it is meant to carry out multi-step operations around a development task.

Local deployment is listed as one of the key features. For teams that need to work with code inside their own environments, running the model locally means data does not have to leave internal infrastructure.

Efficient inference matters for real-world use, too. When a model runs on local hardware, its inference efficiency drives both response speed and hardware cost, which often decides whether self-hosted setups are practical.

JetBrains is best known for its integrated development environments and developer tooling. By open-sourcing its own model, JetBrains signals that it wants to shape not just the tools developers use, but the models underneath them.

So far, the public information centers on the release itself and how the model is positioned. What to watch next is community adoption and how it performs on real coding tasks.