JetBrains releases Mellum2, a 12B MoE coding model
JetBrains published Mellum2 on Hugging Face, positioning the 12B mixture-of-experts model for code-completion and developer-tool workflows.
Read more
JetBrains' Mellum2 is another missed open-model item, and it is directly relevant to developer tools. The company published the model on Hugging Face as a 12B mixture-of-experts release, extending its Mellum line for code-focused workflows. The important detail is distribution: JetBrains is not only embedding AI into IDEs, it is also releasing model artifacts that developers can inspect and use through the broader open model ecosystem. This is a technical release rather than a market-moving headline, but it helps readers track how coding models are fragmenting into specialized, tool-integrated systems. Watch license terms, local serving cost, benchmark behavior inside real IDE tasks, and whether JetBrains routes more of its AI assistant stack through its own models.
Key details: June 1, 2026, JetBrains, Mellum2, 12B mixture-of-experts model, Hugging Face, Code workflows, Developer tools.
Continue swiping for more AI Brief stories.