Back to AIBriefs
LaunchAI Models

JetBrains open-sources Mellum2, a 12B mixture-of-experts model

JetBrains open-sources Mellum2, a 12B mixture-of-experts modelHugging Face Blog

Mellum2 is a 12B-parameter MoE model with 2.5B active parameters per token, released under Apache 2.0. Designed for production AI workflows—routing, sub-agents, and private on-premises deployment—it offers inference time less than half of comparable models. Trained from scratch, it is specialized for code and natural language, not multimodal.

JetBrains/Mellum2-12B-A2.5B-Thinking12 days agoJetBrains
JetBrains/Mellum2-12B-A2.5B-Instruct10 days agoJetBrains
RT @jetbrains: Mellum2 just launched! Experience strong performance across coding and language...5 days agoHugging Face
RT @jetbrains: Mellum started with code completion. Mellum2 is built for more – handling both...12 days agoHugging Face
JetBrains open-sources Mellum2 to go where Claude Code can’t12 days agoPaul Sawers
Jetbrains Mellum 2: a really good and performant model5 days agogcavalcante8808Discuss
Mellum 2 12B A2.5B12 days agoMiddle_Bullfrog_6173Discuss
Mellum2 Goes Open Source: A Fast Model for AI Workflows | The JetBrains AI Blog12 days agodayanrubenDiscuss
12 days ago