JetBrains releases Mellum2.1, an Apache 2.0 open model for coding agents

Archive entry: this covers an event from October 8, 2026, and we published it on November 8, 2026. The details come from sources dated at the time and may have changed since.

On October 8, 2026, JetBrains released Mellum2.1, an updated open model for coding agents, in a blog post by Bulat Salimzianov. It was the next version of the 12B mixture-of-experts model JetBrains open-sourced in June, with 2.5B active parameters.

What changed

• Mellum2.1 was released under the Apache 2.0 license and made available on Hugging Face.
• The architecture was unchanged since version 2; nearly all of the work went into post-training, mainly reinforcement learning.
• Training used millions of sandboxed runs across thousands of environments, so the model could explore a repository, edit files and check its changes.
• JetBrains compared it with Mellum2, Qwen3.5-9B and Gemma 4 E4B and reported the biggest improvement in agentic coding.
• Multi-token prediction made a single request about 1.6 times faster; under heavy load the model served almost twice as many tokens as Qwen3.5-9B.

What it meant for builders

Developers building self-hosted coding agents got a small, permissively licensed model tuned for working inside repositories. JetBrains aimed it at coding agents and at fast sub-agents running on the user's own hardware.

What was still to come

At release, JetBrains listed GGUF builds for llama.cpp, Ollama and LM Studio, and the MTP head for speculative decoding in vLLM, as still to come.

Sources

• Mellum2.1 Gets to Work: A Fast Open Model for Coding Agents (JetBrains, October 8, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.