---
title: "Introducing Mellum2: A 12B Mixture-of-Experts Model by JetBrains"
description: "JetBrains is releasing Mellum2, a 12B-parameter Mixture-of-Experts model trained from scratch on natural language and code. It targets latency-sensitive, high-throughput workloads including routing, retrieval-augmented generation (RAG), summarization, sub-agents, and coding features. The model has 12B total parameters but activates 2.5B per token, with its architecture keeping total capacity high while making inference more efficient for real-time use. Released under the Apache 2.0 license, Mellum2 is described as competitive with similarly sized open models on code generation, reasoning, science, and math benchmarks while delivering more than 2x faster inference. Its text-and-code focus supports high-throughput production workloads and private deployments where proprietary code or internal data may be involved."
---

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

[Hugging Face](https://yomu.fyi/company/hugging-face) · Nikita Pavlichenko · Jun 1, 2026

**Type:** Announcement

## Summary

JetBrains is releasing Mellum2, a 12B-parameter Mixture-of-Experts model trained from scratch on natural language and code. It targets latency-sensitive, high-throughput workloads including routing, retrieval-augmented generation (RAG), summarization, sub-agents, and coding features. The model has 12B total parameters but activates 2.5B per token, with its architecture keeping total capacity high while making inference more efficient for real-time use. Released under the Apache 2.0 license, Mellum2 is described as competitive with similarly sized open models on code generation, reasoning, science, and math benchmarks while delivering more than 2x faster inference. Its text-and-code focus supports high-throughput production workloads and private deployments where proprietary code or internal data may be involved.

## Context

Modern AI systems often use multiple latency-sensitive model calls for routing, retrieval, summarization, planning, validation, and tool use, and many of these operations do not require the largest available model. The release positions Mellum2 as a focused component for making larger AI systems faster, cheaper, and easier to control.

## Approach / What changed

Mellum2 uses a Mixture-of-Experts architecture with 12B total parameters and 2.5B active parameters per token. It was trained from scratch on natural language and code, focuses on text and software engineering rather than multimodal tasks, and is released under the Apache 2.0 license for open and private deployment.

## Takeaways

- Mellum2 activates 2.5B of its 12B parameters per token, using a Mixture-of-Experts design to improve inference efficiency and reduce serving cost for real-time workloads.
- The model is intended for routing and orchestration, RAG context compression and summarization, agent subtasks, and high-throughput coding features.
- The technical report evaluates code generation, reasoning, science, and math; Mellum2 is reported as competitive with similarly sized open models while providing more than 2x faster inference.

**Tags:** [LLMs](https://yomu.fyi/topic/llm), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Open Source](https://yomu.fyi/topic/open-source), [Performance](https://yomu.fyi/topic/performance)

- Source: [Hugging Face](https://huggingface.co/blog/JetBrains/mellum2-launch)
- Source URL: https://huggingface.co/blog/JetBrains/mellum2-launch
- Ingested by Yomu: 2026-08-29T01:49:41.866Z

[Read original post](https://huggingface.co/blog/JetBrains/mellum2-launch)