Skip to content
AI HOT (Curated Pool)

Introducing Mellum2: JetBrains' 12B Mixture-of-Experts Model

介绍 Mellum2:JetBrains 推出的 12B 混合专家模型

JetBrains published a Hugging Face blog post introducing Mellum2, confirming a mixture-of-experts architecture and a 12B parameter scale; the snippet does not disclose training data, license, benchmarks, or deployment conditions.

Why it matters: JetBrains has open-sourced Mellum2, a 12B-parameter mixture-of-experts model that activates only 2.5B parameters per inference and runs more than twice as fast as models of the same size, suited to latency- and throughput-sensitive work.

Read the original ↗Export Markdown