Model Releases

JetBrains releases Mellum2: a 12B-parameter mixture-of-experts model

Original title:Introducing Mellum2: A 12B Mixture-of-Experts Model by JetBrains

AI summary

Mellum2 is a 12B-parameter mixture-of-experts (MoE) model trained from scratch by JetBrains, focused on natural language and code. Each token activates only 2.5B parameters, and inference speed can reach more than 2x that of comparable models, making it suitable for high-throughput, low-latency scenarios. The model supports routing, RAG, summarization, sub-agents and private deployment, and is open-sourced under the Apache 2.0 license. On code generation, reasoning, science and math benchmarks, Mellum2 is competitive with open models of similar scale.

Key Takeaways

  • Mellum2 is a 12B-parameter mixture-of-experts (MoE) model trained from scratch by…
  • Each token activates only 2.
  • 5B parameters,and inference speed can reach more than 2x that of comparable models,making it…
  • The model supports routing,RAG,summarization,sub-agents and private deployment,and is open-sourced…
  • 0 license.

Quick Facts

Published
Updated
Source
Hugging Face 社区博客(混合发现)
Category
Model Releases
AI score
71
Read the original Read the full article at the original source
Why it matters

JetBrains 开源了 Mellum2,一个激活参数仅 2.5B 的 12B MoE 模型,专为代码和问答管道设计的快模型。做实时 RAG 或子代理时,终于不用硬扛大模型了。

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: JetBrains releases Mellum2: a 12B-parameter mixture-of-experts model. June 1, 2026. https://new.nat6.net/en/news/cmpvfn56c06fxsl0zknga1ztr/

View all →