Ok all you model dorks, the day I’ve been awaiting: Mellum 2.1 is out. 16 GB target, coding-first, ethical-ish training, open, European, fast as ****.
Details: 12-A2.5B and if you follow this space, you’ll know the under-32G-laptop target has kind of been abandoned. Certainly by MoE architectures.
How fast? What’s it good for? I have a video script ready, and like 3 more after that.
https://blog.jetbrains.com/ai/2026/10/mellum2-1-gets-to-work-a-fast-open-model-for-coding-agents/