Original sourceJetBrains Blog
Summary
JetBrains released Mellum2.1, a 12B mixture-of-experts model optimized for coding agents and on-premises deployment, with the same architecture as Mellum2, 2.5B active parameters, and released under Apache 2.0 license. This version was post-trained via reinforcement learning in real environments, e…
Key points
- Teams wanting to self-host or run coding agents on-premises can evaluate this new option with a permissive license and smaller size.
- Open-source, self-hostable coding agent models increase enterprise choices under the premise of not leaking data, influencing AI development tool deployment strategies.
- For enterprises valuing data sovereignty, Mellum2.1 offers an agent solution that does not send code to external APIs, but performance and integration costs must be verified independently.
Editorial note
This page is Code & Chain's editorial summary of public sources. It may be prepared with AI assistance and published through an automated workflow. Refer to the original sources; this content is not investment, legal, or tax advice.