Mellum2 Goes Open Source: A Fast Model for AI Workflows (blog.jetbrains.com) AI

JetBrains says it has open-sourced Mellum2, a 12B parameter “focal” language model designed for faster, lower-cost production AI—using a Mixture-of-Experts approach where only part of the model is active per token. The post highlights use cases like routing/orchestration, low-latency RAG summarization, and running private or self-hosted AI workflows, and notes the model is specialized for natural language and code (not multimodal).