← Founder Notes
Archive

The ide lab just shipped its own model. jetbrains released mellum 2.1 on oct 10, a…

Yethikrishna ROriginal on Threads

the ide lab just shipped its own model. jetbrains released mellum 2.1 on oct 10, a 12-billion-parameter moe built for coding agents, joining the open-model rush from a tooling company.

the agents now run on the editor's own weights.

Context

JetBrains' blog says it is releasing Mellum2.1, the next version of its 12B mixture-of-experts model, trained with reinforcement learning in real environments for coding agents and fast sub-agents that run on your own hardware. MarkTechPost says it activates 2.5B parameters per token and ships under Apache 2.0 on Hugging Face.

Traictory notes the architecture is identical to Mellum2, which JetBrains open-sourced in June 2026.

How it compares

The 12 billion parameter mixture-of-experts coding model matches the sources. JetBrains' blog and MarkTechPost date the release Oct 8, not Oct 10; Traictory's write-up is dated Oct 10. The 'joining the open-model rush' framing is the author's reading.

'the agents now run on the editor's own weights' is the author's opinion.

Related work

Watch next

  • Read the model card on Hugging Face for the benchmark table.

Sources

  1. JetBrains Blog, Oct 8, 2026blog.jetbrains.com
  2. MarkTechPost, Oct 8, 2026marktechpost.com
  3. Traictory, Oct 10, 2026traictory.com

Provenance

The note above is reproduced unedited from the original post, first published on Threads on 11 October 2026 at 01:05 IST. Sources are the papers and datasets the note draws on.

View the original post
Embed this note
<iframe src="https://founder.myndlabs.tech/notes/embed/the-ide-lab-just-shipped-its-own-model-DeU2Oc6kbPi" width="480" height="420" style="border:0;max-width:100%" loading="lazy" title="The ide lab just shipped its own model. jetbrains released mellum 2.1 on oct 10, a…"></iframe>

More notes