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

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.

Takeaways

1. 1

   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.

2. 2

   The model is intended for routing and orchestration, RAG context compression and summarization, agent subtasks, and high-throughput coding features.

3. 3

   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.